Winners and Losers from Sovereign Debt Inflows∗

Fernando Broner1, Alberto Martin2, Lorenzo Pandolfi3, and Tomas Williams4

1CREi and Barcelona GSE2European Central Bank, CREi, Barcelona GSE

3University of Naples Federico II and CSEF4George Washington University

June 15, 2020

Abstract

We study the transmission of sovereign debt inflow shocks on domestic firms. We exploitepisodes of large sovereign debt inflows in six emerging countries that are due to the announce-ments of these countries’ inclusion in two major local currency sovereign debt indexes. We showthat these episodes significantly reduce government bond yields and appreciate the domesticcurrency, and have heterogeneous stock market effects on domestic firms. Firms operating intradable industries experience lower abnormal returns than firms in non-tradable industries. Inaddition, financial, government-related, and firms which rely more on external financing expe-rience relatively higher abnormal returns. The effect on financial and government-related firmsis stronger in countries that display larger reductions in government bond yields. The effecton tradable firms is stronger in countries where the domestic currency appreciates more. Weprovide a stylized model that rationalizes these results. Our findings shed novel light on thechannels through which sovereign debt inflows affect firms in emerging economies.

JEL Classification: F31, F32, F36, G15, G23

Keywords: Sovereign debt; capital inflows; exchange rate; government bond yields; stock prices; emerging

markets

∗We are grateful to Graciela Kaminsky, Tommaso Oliviero, Marco Pagano, Giovanni W. Puopolo, Stefano Rossi,Tom Schmitz, Annalisa Scognamiglio, Saverio Simonelli, seminar participants at the Bank of Italy, and participants atthe III Interdisciplinary Sovereign Debt Research and Management Conference (DebtCon3) conference, the XV CSEF-IGIER Symposium on Economics and Institutions, the 2020 Annual Meeting of the American Economic Association,and the XXI Workshop on Quantitative Finance, for several insightful comments and suggestions. We thank ColtonLarson and Genna Tatu for excellent research assistance, and Mariano Cosentino for his help in the preparation of thedataset. We gratefully acknowledge financial support from the Einaudi Institute for Economics and Finance (2017EIEF research grant) and the Columbian College Facilitating Fund from George Washington University. All errorsremain ours.

1

1 Introduction

Financial globalization in emerging economies has increased remarkably over the last two decades.

Loosening of capital controls and investors’ search for yield in times of low interest rates contributed

to an upsurge in the presence of foreign investors in emerging asset markets. Local currency

sovereign debt markets, in particular, experienced an unprecedented rise in international investors’

participation. Since the Global Financial Crisis, the share of emerging countries’ local currency

sovereign debt held by foreign investors more than doubled, going from about 10% in 2009 to

approximately 25% in 2018 (BIS, 2019).

The increased participation of foreigners in domestic debt markets is believed to have significant

effects on the recipient economies. Yet, the exact nature of these effects is still debated. For instance,

small-open economy models with full financial integration leave little or no role for shocks to the

foreign demand for domestic assets.1 But recent studies argue that shocks to global investors’

demand indeed play a role in determining sovereign bond yields and exchange rates.2 Most of these

studies depart from the standard assumption of perfect financial markets by introducing frictions

such as market segmentation and limited arbitrage. Under this view, as long as domestic financial

markets are not perfectly integrated in international markets, an increase in foreign demand for

domestic sovereign debt leads to net capital inflows. These inflows, in turn, affect domestic firms

through their effects on interest and exchange rates.3 These effects are likely to be heterogeneous

depending, for instance, on whether a firm operates in a tradable sector or is financially dependent.

Assessing the channels through which sovereign debt inflows can transmit to domestic firms

is challenging from an empirical point of view. The main reason is that sovereign debt flows are1For example, in models of exchange rates determination based on uncovered interest parity, such as Obstfeld andRogoff (1995), portfolio flows do not affect the equilibrium exchange rates, as noted in Gabaix and Maggiori (2015).

2Gabaix and Maggiori (2015) show theoretically that capital inflows can appreciate the currency of recipient countriesas long as international financial intermediaries have limited risk-absorbing capacity. Du and Schreger (2016) arguethat local currency sovereign debt markets are typically not internationally integrated, so that clientele demandshocks are important determinants of local currency sovereign bond prices.

3Consistent with this, Pandolfi and Williams (2019) provide empirical evidence that sovereign debt inflows increaseboth the value of the domestic currency and the price of local-currency denominated sovereign bonds. Relatedly,Hofmann et al. (2019) find that currency appreciations in emerging economics are often contemporaneous to reductionin government bond yields.

2

clearly endogenous to macroeconomic conditions that affect and are affected by domestic economic

conditions. Typically, sovereign debt flows, sovereign bond yields, exchange rates, and firms’ fun-

damentals are jointly determined. For instance, they can all be affected by positive technological or

political shocks which reduce sovereign risk, improve fundamentals and, at the same time, attract

foreign investors. As a result, any correlation between sovereign debt inflows and the profitability

of domestic firms cannot be interpreted as evidence of a causal relationship.

The aim of this paper is to assess the impact of sovereign debt inflows on domestic firms,

shedding light on the main transmission channels. To do so, we exploit episodes of large sovereign

debt inflow shocks in six emerging countries (Colombia, Czech Republic, Mexico, Nigeria, Romania,

and South Africa). Specifically, we take advantage of sudden and unanticipated announcements of

country inclusions in two major sovereign debt indexes: the Citigroup World Government Bond

Index (WGBI) and the J.P. Morgan Government Bond Index Emerging Markets (GBI-EM). These

two indexes represent two of the most widely tracked benchmarks for international investors in local

currency sovereign debt markets. Because of this, country inclusion announcements induce large

rebalancings in the portfolios of international investors who, in order to replicate the composition

of the index they follow, suddenly increase their demand for the newly included country’s local

currency government bonds and domestic currency.

These announcements provide an ideal setting to address our research question. First, they

trigger large inflows which are specific to the sovereign debt markets of newly included countries:

as the two indexes are exclusively composed of sovereign bonds, their rebalancings only entail

inflows to sovereign bond markets and not to equity nor corporate debt markets. Second, the dates

in which index providers announce the inclusions are not anticipated by investors, nor they coincide

with important news about the economy of included countries or with major policy changes. Thus,

these feature allow us to adopt an event study methodology – which exploits the unexpected timing

of the announcements – to assess the impact of sovereign debt inflows on the cost of government

debt, the domestic currency, and the stock market returns of domestic firms.

Our results are as follows. In the two days following the announcement episodes, 5-year local

3

currency sovereign bond yields in all countries drop significantly. The 2-day reduction in yields

is sizable, as it corresponds on average to 32.5 basis points. Exchange rates also move in the 2

days following the announcement: the domestic currency appreciates in all countries, the average

appreciation being 1.1 percentage points.

Domestic firms’ stock prices also respond significantly and heterogeneously to the announce-

ments. Figure 1 provides a graphical representation of these results. The price of financial and

government-related firms tend to increase sharply in the first trading day following the announce-

ment episodes. Instead, the market value of firms operating in tradable sectors remain relatively

stable when inclusions are announced. Importantly, while prices of all these firms are on the same

trend in the 7 trading days prior to the episodes, they clearly diverge in the dates of the announce-

ments. Further, this divergence persists even 20 trading days after the announcements.

To assess the economic magnitude of these heterogeneous effects, we analyze how abnormal

returns of firms evolve after each announcement episode in a series of regressions in which we can

control for potential overlap between different firm categories and for firms’ dependance on exter-

nal financing. These regressions show that financial and government-related firms experience larger

than average cumulative abnormal returns (CARs) in the two days following the announcements.

Instead, tradable firms experience relatively lower CARs in the aftermath of the events. Addi-

tionally, consistent with the hypothesis of a potential pass-through from financial to financially-

constrained firms, we find that CARs tend to be larger for firms operating in more financially

constrained industries.

We conjecture that these effects of sovereign debt inflows are mostly driven by changes in

sovereign bond yields and exchange rates. Indeed, we find that countries which experience larger

reductions in the 5-year local currency sovereign bond yield are the ones in which the effect on

financial and government-related firms is more pronounced. Similarly, countries whose currency

appreciates more are the ones where tradable firms are more affected by the announcement. Quan-

titatively, sovereign debt inflows that lead to a 100 basis points reduction in the 5-year government

bond yield, increase the CARs of financial and government related firms by 1.7 percentage points.

4

In turn, inflows that lead to a 1% appreciation of the exchange rate are associated with 0.41 per-

centage points lower CARs for tradable firms. We also find that the cross-country heterogeneity

in the 2-day changes in sovereign bond yields and exchange rates following each announcement is

related to the magnitude of the inclusion-driven sovereign debt inflows in each country. We provide

additional results which support to our empirical analysis and various robustness checks.4

We present a stylized model to formalize our empirical findings. The model analyzes a small-

open economy with tradable firms, non-tradable firms and banks. The tradable sector is capital

intensive. Banks hold government debt and intermediate between domestic savers and borrow-

ers. Crucially, there are no domestic financial frictions but there is limited borrowing from the

international financial market.

We model sovereign inflows as an increase in foreign purchases of government bonds. These

inflows are shown to affect the economy through two channels. First, since the economy is con-

strained, they entail net capital inflows and thus reduce the domestic interest rate. This raises the

price of government debt, which benefits banks directly, and expands domestic credit and thus the

relative production of the tradable good. Among all firms, moreover, lower interest rates benefit

especially those that are financially constrained, and – by reducing the cost of financing for the

government – also those that are related to the public sector. Second, sovereign inflows raise the

relative supply of tradables in the domestic economy, thereby appreciating the real exchange rate

and benefitting the non-tradable sector. The model thus illustrates the basic forces at work behind

our empirical estimates.

Our results are mostly related to three broad strands of the literature. First, our study is related

to the growing literature that focuses on the effects of financial flows on government bond yields

and exchange rates in imperfectly integrated financial markets. Among those, for instance, Gabaix

and Maggiori (2015) develops a model of exchange rate determination in which global investors’4For instance, we show that the difference in the post-announcement CARs of financial and government-relatedfirms vs. tradable firms is not driven by different pre-existing trends. Further, consistent with the fact that theannouncements entail shocks which are specific to countries’ sovereign debt markets, we show that sovereign debtinflows in these countries increase sharply in the announcement year, while private inflows do not.

5

portfolio flows can lead to currency appreciations in recipient countries because of the limited

capacity of financiers to absorb currency demand shocks. Evidence of the imperfect integration

across local currency sovereign debt markets is provided in Du and Schreger (2016), which argues

that an increased presence of foreign investors in local currency sovereign debt markets can have

important implications for the pricing of local-currency denominated bonds.5 Relatedly, Pandolfi

and Williams (2019) presents evidence that sovereign debt inflows due to international investors’

portfolio rebalancings can increase the price of local currency sovereign bonds and appreciate the

domestic exchange rate. More broadly, our results are also related to the vast literature that

focuses on the effects of demand shocks on sovereign bond yields in both emerging and developed

economies.6 Very related to ours is the study by Kolasa and Wesolowski (2020), which develops a

two-country open economy model with segmented markets, and show that quantitative easing in

developed countries can induce sovereign debt inflows in emerging countries which boost domestic

demand but dampen international competitiveness. Our results are indeed consistent with these

theoretical predictions.

Our results speak also to a second important strand of literature that analyzes how changes

in government bond yields and exchange rates affect firms. Many of these studies focus on the

stock market effects of increased sovereign default risk on domestic firms, and the consequent

rise in sovereign bond yields on domestic firms.7 Our findings are in line with the evidence in

these studies even though we exploit events which reduce government financing costs, rather than

increase them. They are consistent also with the theoretical predictions in Gennaioli et al. (2014)

– according to which changes in sovereign default risk and sovereign bond yields can transmit to5Other studies which focus on the interactions between international investors’ participation and the pricing of localcurrency sovereign bonds are those by Borri and Shakhnov (2018), Hofmann et al. (2019), and Morelli et al. (2019).Relatedly, Warnock and Warnock (2009) and Kohn (2015) analyze the effects of foreign purchases of U.S. governmentdebt on U.S. sovereign bond yields.

6Among others, Vayanos and Vila (2009), and Greenwood and Vayanos (2010) discuss the role of demand shocksfrom preferred-habitat investors in the determination of government bond yields. Recent studies focusing on thequantitative easing measures taken after the Global Financial Crisis also highlight that demand shocks and marketsegmentation can be important determinants of government bond yields (Krishnamurthy and Vissing-Jorgensen,2011; Krishnamurthy et al., 2017).

7For instance, Hébert and Schreger (2017) estimates the cost of sovereign default for listed Argentinian companiesexploiting legal rulings against Argentina. Similarly, Andrade and Chhaochharia (2018) and Chari et al. (2018)analyze the costs of sovereign default (in Europe and Puerto Rico, respectively) and find that firms that are moreclosely related with the domestic government tend to be more sensitive to changes in the risk of default.

6

firms through the balance sheet of banks holding government bonds8 – and Arellano et al. (2017)

– according to which firms with larger external financing needs are more sensitive to changes in

sovereign default risk. Additionally, our findings are related to numerous studies analyzing how

exchange rate movements affect the terms of trade of a country and, therefore, domestic firms.9

Finally, this paper also relates to a vast international macroeconomics literature that focuses on

the channels through which capital flows affect firms in the economy.10 This literature mostly focus

on FDI, bank flows, and equity portfolio flows. A study which is closely related to ours is Blanchard

et al. (2017), which argues that, while equity inflows can be expansionary for emerging countries,

bond inflows should be in general contractionary. We show that, when taking into account the

effect on government bond yields, sovereign debt inflow shocks can have both expansionary and

contractionary effects, depending on the sectoral composition of the economy and the sensitivity of

domestic firms to changes in sovereign bond yields and the exchange rate. Finally, our results are in

line with the evidence in Benigno et al. (2015) of a systematic reallocation of labor and investment

from tradable to non-tradable sectors, after episodes of large capital inflows.

The rest of the paper is structured as follows. Section 2 presents the empirical setting, decribing

the testable implications that we bring to data, the announcement episodes, and the data used to

conduct the analysis. Section 3 presents the first set of results on sovereign bond yields and exchange

rates. Section 4 presents the main results on the stock market effects of sovereign debt inflow shocks

on domestic firms. Section 5 discusses the potential determinants of the heterogenous cross-country

reaction in response to the announcement episodes, and the medium-long term consequences of

sovereign debt inflows on the economy of recipient countries. Section 6 presents a model that

rationalizes our empirical findings. Finally, Section 7 concludes.8Altavilla et al. (2017), and Bottero et al. (2020) also provide evidence which is consistent with this channel.9Gabaix and Maggiori (2015) argue that a depreciation of the domestic currency improves a country’s terms of trade,which in turn leads to an increase in employment in tradable industries. This view is questioned by other studiesaccording to which exchange rate depreciations may not affect firms’ competitiveness as export prices are sticky ina dominant currency (i.e., the U.S. dollar) (Gopinath et al., 2020).

10See among others Schnabl (2012), Alfaro et al. (2014), Lane and McQuade (2014), Baskaya et al. (2017), andCalomiris et al. (2020). Related to this strand of literature are the studies on capital account liberalizations andtheir effects on firms’ stock market value (see Chari and Henry (2004), Chari and Blair Henry (2008) and Larrain(2015) among others).

7

2 Empirical Setting

2.1 Conceptual Background

In the textbook small, open economy model without frictions, shocks to foreigners’ demand for

local currency sovereign debt should have no impact on the domestic economy. In a nutshell, the

gross inflows entailed by foreigners’ demand for domestic debt would be compensated one-for-one

by a reduction of other gross inflows or by an increase of gross outflows. In the presence of financial

frictions, however, this is not true.

First, financial frictions may hamper access to international financial markets. In this case,

sovereign inflows increase the availability of financial resources and lead to a reduction in gov-

ernment yields and other domestic interest rates. In other words, sovereign inflows reduce the

government’s reliance on domestic financial markets and free up resources for the private sector

(Becker and Ivashina, 2018; Ongena et al., 2019; Williams, 2018).11 In this way, they foster pri-

vate investment and fuel the expansion of firms and sectors that rely heavily on external finance

(Arellano et al., 2017). Moreover, through their effect on the price of public debt, sovereign in-

flows provide an windfall for bondholders, in particular domestic financial institutions (Gennaioli

et al., 2014). We will test for these implications in the data by empirically analyzing the effects of

sovereign inflows on domestic financial firms, on firms that are more financially constrained, and

on firms that are connected to the government. (Chari et al., 2018).

Second, by increasing the relative availability of tradable goods, sovereign debt inflows should

lead to an appreciation of the exchange rate. This appreciation should in principle hurt, in relative

terms, firms in the tradable sector (Gabaix and Maggiori, 2015).

In Section 6, we formalize this conceptual discussion with the help of a stylized model. For

now, we use it to guide our empirical analysis. In particular, we analyze the effects of exogenous11In other words, foreign purchases of public debt reduce its “crowding-out” effect thereby raising private investment(see Broner et al. (2014), Broner et al. (2019), and Priftis and Zimic (2020)).

8

shocks to the foreign demand for local currency sovereign debt on the stock returns of: (i) domestic

financial firms; (ii) firms that are connected to the government ; (iii) firms that are more financially

dependent; and (iv) tradable firms. To the extent that sovereign inflows relax domestic financial

conditions, we expect (i), (ii), and (iii) to be positive. To the extent that sovereign inflows lead to

an appreciation of the exchange rate, we expect (iv) to be negative.

2.2 Index Inclusions as Sovereign Debt Inflow Shocks

In order to analyze the consequences of sovereign debt inflows for domestic firms, we exploit episodes

of country inclusion in two major global local currency sovereign debt indexes which are widely

used as benchmarks by international investors: the Citigroup WGBI and the JP Morgan GBI-

EM.12 Specifically, the episodes we exploit are the inclusions in the GBI-EM of local-currency

denominated sovereign bonds issued by the governments of Colombia, Czech Republic, Nigeria,

and Romania, and the inclusions in the WGBI of Mexican and Southern African local-currency

denominated sovereign bonds. These episodes were announced by index providers on specific dates

which we retrieved from their websites.13

These inclusion events constitute an ideal laboratory to address our research question as they

trigger massive sovereign debt inflows by foreign investors. International investors who are bench-

marked against each index have indeed the incentive to rebalance their portfolios, purchasing

sovereign bonds of the newly included country to replicate the index composition.14 Even though

the inclusion is usually implemented gradually by index providers over the 3 to 12 months sub-

sequent to the announcement, more active investors have the incentive to start rebalancing their12For more details on how these indexes are created and their importance see Section A.1 in the Appendix.13In particular, the first trading days after the announcement episodes are the 19th of March 2014 (Colombia), the22nd of February 2017 (Czech Republic), 31st of March 2010 (Mexico) the 16th of August 2012 (Nigeria), the 16thof January 2013 (Romania), and the 17th of April 2012 (South Africa). More details about these two indexes areprovided in Section A.2 in the Appendix.

14Evidence of international investors’ tendency to replicate the composition of the indexes they track is abundant inthe literature. See for instance Cremers et al. (2016) and Raddatz et al. (2017). Basak and Pavlova (2013) andKashyap et al. (2018) show theoretically that asset managers whose performance are evaluated against a commonbenchmark have the incentive to increase their demand for securities included in the benchmark index. This canin turn affect the price of firms in the benchmark, which end up being subsidized by asset managers.

9

portfolios already at the announcement date in order to minimize the cost of the rebalancing. Fig-

ure 2 depicts the evolution in the average share of sovereign debt held by private foreign investors

across 4 of the 6 countries in our sample around the corresponding announcement dates.15 The

figure clearly shows that, already in quarter of the announcement, the foreign investor base of these

countries’ sovereign debt increases sharply and deviates from the pre-announcement trend.

To grasp the magnitude of these inclusion-driven shocks, we follow Pandolfi and Williams (2019)

and adapt their Flows Implied by Rebalancings (FIR) measure to our setting. We compute the

estimated inflow shocks to the sovereign debt market of each country as the change in benchmark

weight – calculated over the entire implementation period – multiplied by the assets under man-

agement of funds tracking their returns against the corresponding benchmark index, normalized

by the market value of the securities that are going to be included in the index. The FIR measure

captures the total inflows that would enter the country if all institutional investors were to pre-

cisely replicate the index composition. For the episodes in our sample the estimated FIR is 18%

in Colombia, 25% in Czech Republic, 12% in Mexico, 31% in Nigeria, 30% in Romania, and 10%

in South Africa. Hence, the estimated sovereign debt inflows due to the inclusion events are very

large, especially if compared to the size of these countries’ sovereign debt markets.

Additionally, these events share some features which are key for our identification strategy.

First, the dates in which index providers announce the inclusions are unexpected. Of course,

the inclusions themselves do not come as a surprise, as markets in most cases were expecting

these countries to get included into an emerging sovereign debt index at some point in the future.

However, the exact timing of the inclusion in the index is not anticipated by investors.16 Second,

the announcements made by index providers are not contemporaneous to macroeconomic shocks

nor to relevant policy changes which might have a direct effect on stock prices.17 Because of15The data used to produce this figure is from Arslanalp and Tsuda (2014), and do not include information of foreignprivate holdings of Czech and Nigerian debt.

16Consistent with the unexpected nature of the shock, in Section 3, we show that 5-year government bond yieldsin all newly included countries drop significantly on the exact dates in which inclusions are announced, whichmakes implausible that investors were anticipating the events and trading accordingly in the days prior to theannouncement dates.

17Precisely because of this, the inclusion of Argentina in the GBI-EM is not part of our sample. For this country,the inclusion event coincides with the removal of capital controls, which mostly likely had a direct impact on the

10

this, the announcements are very unlikely to reveal information about changes in the countries’

fundamentals. If that was the case, the inclusion episodes should be followed by massive inflows

into the private sector as well.

The evidence in Figure 3 shows that this is not the case. In it, we depict the cross-country

average public and private inflows in the three years before the announcement episode, and the

average private and public inflows in the year of the inclusion event. In the figure, private inflows

are made of portfolio debt, equity and foreign direct investment inflows, while public inflows are

inflows to the sovereign debt market. The figure shows that the events trigger large inflows which are

specific to the sovereign debt markets of newly included countries: public inflows nearly triple in the

year of the inclusion episode, compared to average inflows in the three years before. Instead, private

inflows remain almost unchanged. This evidence provides important support to our empirical

strategy, as it further corroborates the hypothesis that the announcement episodes have a negligible

information content and are not associated to other events which trigger also private inflows to the

countries.18

2.3 Data

To conduct the empirical analysis, we combine data from several sources. We collect from Datas-

tream the time-series of daily (end-of-day) stock prices of domestic public companies in Colombia,

Czech Republic, Mexico, Nigeria, Romania, and South Africa. We gather from Datastream also

additional information about companies in these countries, including the International Securities

Identification Number (ISIN), the industry classification, and a concise description of each firm’s

business activity. We combine this information with end-of-year balance sheet data from World-

scope.

We then collect the daily time series of local currency 5-year government bond yields from

Argentinian equity market.18Figure A1 in the Appendix reports the evolution of private and public inflows around the announcement year foreach of the countries in our sample, separately.

11

Bloomberg, and the daily time series of (end-of-day) exchange rates from Datastream. The exchange

rate is computed as the amount of local currency needed to buy one U.S. dollar. To compute the

abnormal returns of firms in our sample, we gather data about regional and global factors: from

Datastream we obtain the daily time series of returns on the MSCI Emerging Markets Index and

the MSCI World Index.

We supplement these data with additional firm-specific information. In particular, to identify

firms which are closely related to the domestic government, we proceed in two steps. First, from

Thomson Reuters Securities Data Company (SDC) Platinum we retrieve the list of firms whose

major shareholder is the state. Second, we perform a search in the business description of firms and

look for the words government or public. Hence, we construct the indicator variable 1(Govt), which

equals one if a company is partially owned by the domestic government or its business activity is

related to the domestic government. To identify financial firms, we follow the industry classification

in Datastream. Specifically, we construct the indicator variable 1(Financial) which equals to 1 if

a firm is classified as a bank, a financial firm, an investment firm, or a life insurance company. We

use the industry classification in Datastream also to identify firms operating in tradable sectors. In

particular, we construct the indicator variable 1(Tradable) which takes value 1 if a firm operates

in a tradable industry and 0 otherwise, following the classification scheme in Mian and Sufi (2014).

Further, we use the balance sheet information from Worldscope to measure firms’ dependance on

external financing. We follow Rajan and Zingales (1998) and compute, for each firm, the ratio

between capital expenditures net of cash flows from operations, and capital expenditures. We then

compute the median of this measure in each industry and use it as our proxy for firms’ external

financial dependance (which we denote as EFD).

To identify firms which have access to corporate debt markets, we retrieve from Thomson

Reuters SDC Platinum the list of companies which issue corporate bonds or have syndicated loans

in our sample of countries. We match firms in this list to firms in our database using the SEDOL

and generate the dummy variable 1(DebtIssuer) which equals 1 in case of a successful merge.

Finally, we merge the list of firms in our sample with the list of companies which are included

12

in the MSCI Emerging Markets index, that we get from the MSCI website. Since stocks of these

companies are more likely to be held by foreigners, we use this information to create the indicator

variable 1(ForOwnership), which is equal to 1 if a company is included in this equity index in the

quarter preceding the shock and 0 otherwise.

Table 1 reports summary statistics about firms in our sample. The table reports the share of

financial firms, government-related firms, and tradable firms (on the announcement date) in the

entire sample and in each country, separately. About one third of firms in the sample operate in

tradable sectors, while 14% of firms are financial firms and 9% of them are classified as government-

related. The country where the financial sector is most prominent is Colombia, where 36% of firms

in our sample are financial firms. Czech Republic and Romania, instead, feature the largest share

of government-related firms and tradable firms, respectively (36% and 54%). In all countries, there

are firms which are included in the MSCI Emerging Markets index, and firms issuing corporate

bonds or syndicated loans. Finally, the least represented country in our sample is Czech Republic,

with 14 companies, while the most represented one is South Africa, with 361 companies. In total,

our sample is composed of 909 companies.19

3 Sovereign Debt Inflows, Exchange Rates, and Sovereign Bond

Yields

According to our hypotheses, sovereign debt inflows can play an important role in the determination

of sovereign bond yields and exchange rates in recipient countries. In this section, we therefore start

our empirical analysis by assessing the response of these two variables to the country inclusion

announcement episodes in our sample.

First, in Figure 4, we depict the daily time series of each country’s local currency 5-year gov-19In our sample, we consider each observation as a distinct company. Actually, 28 companies appear twice in ourdataset, as each of them issue two distinct securities. In the Appendix, we show that reducing each of thesecompanies to a single observation does not change the results of our analysis.

13

ernment bond yield in the 2 years around the announcement episode. The figure clearly shows

that government yields drop sharply when index providers announce the inclusion of countries in

their benchmark indexes. Further, the figure shows that there are no common pre-announcement

trends in government bond yields across countries, thus supporting our identifying assumption that

announcement dates were unexpected.

Figure 5 reports the daily time series of each country’s exchange rate in the 2 years around the

announcement episodes. The exchange rate in each country is the amount of domestic currency

needed to buy one US dollar and is normalized to its value in the last trading day before the an-

nouncement. In all countries, the exchange rate drops on the announcement date, which means that

the domestic currency appreciates in response to the inclusion announcements. The appreciation is

not as sharp as the drop in the yields, which can be explained by the fact that exchange rates are

much more volatile than government bond yields, even in calm periods. What is important to note

is that, as for the sovereign bond yields, there is not a pre-announcement trend which is common

to all countries.

To assess the economic magnitude of these two effects, we compare, for each country, the 2-

day changes in the sovereign bond yield and the exchange rate after the announcement with the

average 2-day changes in the 2 years around the announcement. Figures 6 and 7 provide a graphical

representation of this comparison. In all countries, the 2-day change in both the government bond

yield and the exchange rate after the announcement is larger (in absolute value) than the average

2-day change. In most cases, the 2-day change after the announcement lies in the very left tail of

the distribution of 2-day changes, thus representing an outlier observation. In Table 2, we report

the size of these post-announcement changes. In the two days after the announcement episode,

the local currency 5-year government bond yield decreases, on average, by 32.5 basis points. The

country experiencing the largest drop in the yield is Romania, where the yield decreases by almost

90 basis points. Conversely, the country experiencing the lowest reduction in the yield is Mexico (4.8

basis points). In all countries, the 2-day changes after the announcement episodes are statistically

different from the average 2-day change in the yield, which is close to 0. Similarly, the average

14

appreciation across countries is 1.09 percentage points. The lowest appreciation is the one that

occurs in Nigeria (0.057 percentage points), while the largest ones are observed in Colombia and

Romania (2 percentage points). In all countries, the 2-day appreciation following the announcement

date is statistically different from the average 2-day change in the 2 years around the announcement.

4 The Effects of Sovereign Debt Inflows on Domestic Firms: an

Event Study

To measure the effect of sovereign debt inflow shocks on domestic firms, we conduct a multiple event

study around the dates in which country inclusions are announced by index providers. Specifically,

we calculate the cumulative abnormal returns of domestic listed firms in the two trading days fol-

lowing the announcements – which should reflect changes in firms’ profitability and future prospects

– and use them to test the empirical predictions developed in Section 2.1.

We compute abnormal returns for listed domestic firms in four ways. First, we compute, for

each firm, daily abnormal returns as the difference between the return in each trading day and the

average daily return in the year preceding the announcement.20 We then cumulate such abnormal

returns on the first two trading days after the announcements and define the resulting CAR as

CARDemeanedi . Second, we compute each firms’ daily abnormal returns as the difference between

the actual returns and the returns predicted by a 1-factor model where the only risk factor is the

daily return of the MSCI EM Index.21 We label the 2-day CAR thus calculated as CAR1-Factori .22

Third, we calculate daily abnormal returns as the difference between the actual returns and the

returns predicted by a 2-factor model which, on top of the regional factor, includes also a global20In particular, the average pre-announcement daily return is calculated over a 252-trading-day window which ends10 days before each country’s announcement date.

21We do not use domestic stock market indexes to compute predicted returns since, in some of the countries in oursample, these indexes are made of few large firms (often financial firms), whose returns have a major impact on theperformance of the index. Thus, the returns of the domestic index would absorb a large part of the variation thatwe want to exploit.

22Factor loadings of the 1-factor model – as well as those of the 2-factor and 5-factor models – are estimated byrunning a series of firm-specific regressions over a time window of 252 trading days which ends 10 trading daysbefore the announcement date in each country.

15

factor, i.e., the daily return of the MSCI World Index. Cumulating these abnormal returns on

the days following the announcement we get the CAR2-Factori . Fourth, we calculate daily abnormal

returns as the difference between the actual returns and the returns predicted by a 5-factor model

which adds to the regional and global factors, a small minus big (SMB), high minus low (HML),

and momentum (WML) factor, which we use to obtain to calculate the CAR5-Factori .23 In our

baseline specification we use 2-day CARs calculated with these four alternative models as our main

dependent variables.24

4.1 Overall Effect of Sovereign Debt Inflow Shocks

We start our empirical analysis by looking at the overall effect of the announcements and the

associated sovereign debt inflow shocks on domestic firms. To this end, we calculate the average

CAR of all firms in our sample in the two days following the announcement dates as:

CARj =

∑Ii CAR

ji

I(1)

where I the total number of observations and CARji is the 2-day CAR of firm i calculated using

model j.

Table 3, Panel A, reports the average post-announcement CAR of domestic listed firms in our

sample. On average, the post-announcement 2-day CARs are positive as they range between 0.11

and 0.22 percentage points, depending on the way abnormal returns are calculated. However,

except for those computed with the first model (the demeaned returns), the average 2-day CAR

is not statistically distinguishable from 0. Panel B of Table 3 reports, instead, the average CAR

before the announcement date, calculated as the average CAR in the interval [t− 3, t− 2], where t

is the first trading day after the announcement. In this case, the average CAR is very close to 0,23To compute the SMB, HML, and WML factors we follow Cakici et al. (2013). These three factors are computedusing stock market data about all countries in our sample.

24In our baseline specification we use 2-day CARs, as most of the event studies which exploit stock market data (seefor instance Alfaro et al. (2017)). However, in the Appendix (Figure A2), we show that results are robust to using3-day, 4-day, and 5-day CARs. Also, in our baseline specification, we exclude observation in the top and bottompercentiles of the country-specific distribution of 2-day CARs, to control for outlier observations.

16

especially if computed using the 1-, 2-, and 5-factor models. In columns 2 to 4, the average pre-

announcement CAR is below 0.05 percentage points. This evidence lends important support to our

empirical strategy, as it corroborates the hypothesis that announcement dates are not anticipated

by investors and shows that, in the days prior to the events, the factor models used to predict

returns deliver a good approximation of the actual returns.

The evidence in Table 3 thus illustrate two important results. First, prior to the announcements,

firms’ abnormal returns are on average very close to 0. Second, in the two days following the

announcements, the average CAR turns positive, but is relatively small in magnitude and not

statistically distinguishable from 0. However, this result actually masks important heterogeneities

across firms, as we show in the next section.

4.2 Heterogeneous Effects of Sovereign Debt Inflow Shocks

As discussed in Section 2.1, sovereign debt inflows can have heterogenous effects on domestic firms

operating in different industries. In particular, financial and government-related firms can benefit

from the inflows, while firms operating in tradable sectors can be relatively hampered by the inflows.

Hence, the small and not significant overall effect of the announcement episodes on domestic firms

might actually be the result of two forces going in opposite directions.

Figure 8 provides evidence that this is indeed the case for firms in our sample. In the figure,

we plot the abnormal returns of firms – cumulated on an interval that starts 3 trading days before

the announcement date in each country and ends 5 days after it – for all firms in our sample (left

panel), and for financial and government-related firms vs. tradable firms (right panel). While the

average aggregate CAR is close to 0 over the entire interval of time, the average CAR of financial

and government-related firms, and that of tradable firms are not. The average CARs of these two

groups of firms are both close to 0 in the days before the announcement, but then sharply diverge on

the announcement date. In particular, while the average CAR of financial and government-related

firms turns positive in the post-announcement period, that of tradable firms becomes negative.

17

Table 4 reports the average post-announcement 2-day CAR of financial firms, government-

related firms, and tradable firms, separately. In the 2 days after the announcement, financial firms

experience positive CARs which are on average equal to 0.65 percentage points (and statistically

different from 0 in all specifications). Similarly, government-related firms exhibit positive CARs of

approximately 0.8-0.9 percentage points (statistically different from 0 under all models). Conversely,

the average CAR of tradable firms is negative but not distinguishable from zero.

Estimates in Table 4, however, do not take into account the potential overlap between the

different groups of firms. For instance, some tradable firms might also be government-related.

Additionally, they do not take into account the heterogeneity of firms in terms of external financial

needs. To take this into account, we therefore estimate the following equation:

CARi = α+ β11(Financial)i + β21(Govt)i + β31(Tradable)i + β4EFDi + εi (2)

where: 1(Financial) is an indicator variable which takes value 1 for financial firms, and 0 otherwise;

1(Govt) is an indicator variable which takes value 1 for firms which are connected to the domestic

government; 1(Tradable) is an indicator variable which takes value 1 for firms operating in tradable

industries, and 0 otherwise; EFD is a measure of firms’ external financial dependance (computed at

the industry level following Rajan and Zingales (1998)); and ε is the error term. This specification

allows us to measure the differential impact of sovereign debt inflows on each of the above-defined

categories of firms, and to test whether the inflows tend to benefit more financially constrained

firms.

Results are reported in Table 5. Financial and government-related firms experience larger than

average CARs in the two days following the announcement episode. The estimated coefficients of

1(Financial) ranges between 0.53 and 0.61, and is statistically significant in all models. Similarly,

the coefficient of 1(Govt) ranges between 0.6 and 0.7, and is always statistically different from 0.

Instead, tradable firms experience significantly lower than average CARs in the 2 days following the

announcement. The estimated coefficients of 1(Tradable) ranges between -0.45 and -0.51, and is

statistically significant in all specifications. Table 5 also reports the difference between the estimated

18

coefficients on financial firms and tradable firms, as well as the difference between the coefficient on

government-related and tradable firms. In all models, these differences are statistically significant

at the 99% confidence level and quantitatively close to 1 percentage point.

Consistent with the hypothesis that more financially constrained firms also benefit from the

shocks, we find that the coefficient of EFD is positive and significant under all specifications.

Firms operating in industries which depend more on external financing experience larger CARs

in the aftermath of the announcements. Quantitatively, a one-standard deviation increase in the

proxy for external financial dependence increases the 2-day CAR by approximately 0.22 percentage

points.25

We then re-estimate the coefficients in Equation 2 controlling for other firm-specific observables.

In particular, we use as additional controls: i) the indicator variable 1(IssueDebt) which takes value

1 for companies that have access to the corporate debt market (either through corporate bonds or

syndicated loans); ii) the indicator variable 1(ForOwnership), which takes value 1 if a stock is

also included in the MSCI Emerging Markets Index, and is therefore more likely held by foreigners;

iii) size, measured by the logarithm of total assets. In principle, some of these controls might be

related to the effect of the announcement on CARs. For instance, firms issuing bonds could benefit

from reductions in government bond yields, as highlighted in Dittmar (2008). At the same time,

these firms are more likely to be the least financially constrained (within each industry), so that

the overall effect is a priori ambiguous.

Table 6 reports the results from this analysis. None of the additional controls have coefficients25In Table A1 in the Appendix, we show that our main results are robust to: i) excluding illiquid stocks, that isstocks whose price never change in the 20 trading days around the announcement date; ii) reducing the weight offirms with more than one stock, by considering for each of these companies only the average 2-day CAR of thecompany’s traded securities; iii) excluding Czech Republic and Nigeria from our sample. Czech Republic is indeedthe least represented country in our sample, while Nigeria is the only country in our sample whose announcementdate is in part ambiguous, as it is unclear whether the news were released on the 14th or on the 15th of August2012 (as we explain in greater detail in Section A.2 in the Appendix). In all cases, our results are quantitativelyand qualitatively very close to those in Table 5. Further, in Table A2 we show that our results are robust tocontrolling for country fixed effects. Finally, in Table A3, Panel A, we replicate our estimates in the days prior tothe announcement, using as dependent variable the CARs in the interval [t − 3, t − 2], where t is the first tradingday after the announcement. None of the 16 coefficients of interest is statistically significant at the 95% confidencelevel, and only one is statistically different from 0 at the 90% confidence level.

19

which are statistically significant. More importantly, including them does not alter our main results,

as our main coefficients of interest remain quantitatively close to those in Table 5. When controlling

for the log of assets, some coefficient fall below the conventional significance threshold, but this

is most likely because balance sheet data are not available for almost 200 firms and therefore

controlling for assets reduces sample size and power. Nevertheless, the difference between the

coefficients on financial and tradable firms, as well as the difference between the coefficients on

government-related and tradable firms, remain large and highly statistically significant under all

specifications.

4.3 Exploiting Cross-Country Variation

Our results so far show that financial and government-related firms tend to have larger than average

CARs after the announcement episodes, while firms operating in tradable industries exhibit lower

than average CARs. These findings do not exploit the cross-country variation in the changes in

sovereign bond yields and exchange rates after the announcement dates (described in Section 3).

Such heterogeneity is potentially important, as it can be exploited to further validate the hypotheses

developed in Section 2.1. According to our theoretical background, the stock market effects should

be related, across countries, to the size of the changes in sovereign bond yields and exchange rates.

In particular, we should find that: i) the effect of the shock on financial and government-related

firms is more pronounced in countries with larger reductions in sovereign bond yields; and ii) the

effect on tradable firms is stronger in countries where the domestic currency appreciates more.

Here, we present additional evidence which is consistent with these two hypotheses.

We first estimate six separate regressions, one for each country, of the form: CARi = α +

β11(Govt&Fin)i + β21(Tradable)i + β3EFDi + εi, where 1(Govt&Fin) is an indicator variable

which takes value 1 for financial and government-related firms. Then, in Figure 9, we plot the esti-

mated β1 coefficient as a function of the country-specific 2-day change in the 5-year local currency

sovereign bond yield (left panel) and the estimated β2 as a function of the country-specific 2-day

currency appreciation. The figure shows that the magnitude of these changes are related with the

20

stock market impact of the announcements. In particular, financial and government-related firms

located in countries where the yield drops more in response to the announcement, experience larger

CARs than firms located in countries where the 2-day reduction in the yield is smaller. More-

over, tradable firms in countries that experience larger domestic currency appreciations have 2-day

post-announcement CARs that are smaller than those of tradable firms in countries with milder

appreciations.26

To quantitatively assess the relationship between the shocks to government bond yields and

exchange rates, and the CARs of firms around the announcements, we estimate the following

equation:

CARi = θc + β11(Govt&Fin)i ×∆Y ieldc + β21(Tradable)i ×%∆ExchRatec + β3EFDi + εi (3)

where: θc are country fixed effects, and 1(Govt&Fin) and 1(Tradable) are interacted, respectively,

with the country-specific 2-day change in the 5-year local currency sovereign bond yield (∆Y ield)

and the 2-day currency appreciation (%∆ExchRate). Both shocks are in absolute values: larger

values correspond to larger reductions in sovereign bond yields and larger appreciations. Under this

specification, the β1 coefficient should capture the cross-country relationship between the reduction

in the sovereign bond yield and the CARs of financial and government-related firms. Similarly, the

β2 coefficient captures the relationship between the appreciation of the domestic currency and the

CARs of tradable firms.

Table 7 reports the estimated coefficient from Equation (3). The coefficient of both interaction

terms are statistically significant and consistent with the above-stated hypotheses. This analysis

allows us also to provide a more quantitative interpretation to our results: according to the estimates

in Table 7, a sovereign debt inflow shock which reduces the 5-year government bond yield by 100

basis points increases the CARs of financial and government-related firms by 1.7 percentage points.

Conversely, a sovereign debt inflow shock which leads to a 1% appreciation of the domestic currency26In Section 5, we discuss some potential determinants of the differential effects of the inclusion announcements onsovereign bond yields and exchange rates across countries.

21

reduces the CARs of tradable firms by 0.41 percentage points.27

5 Discussion

5.1 Estimated Inflows and Heterogeneity of Shocks

The evidence in Section 4.3 shows that the effects of the announcement episodes on the stock prices

of domestic firms are related to the size of the 2-day changes in sovereign bond yields and exchange

rates in the aftermath of the announcements. In this section, we test whether such heterogeneity

is related to the magnitude of the inflows which are expected to enter each of the countries in our

sample as a result of the index inclusion. First, we compute the estimated flows, in U.S. dollars,

entering each country as the change in benchmark weight – over the entire implementation period –

multiplied by the assets under management of funds benchmarked against the corresponding index.

We normalize the inflows first with respect to the total value, in U.S. dollars, of the local currency

sovereign bonds which are going to be included in the index, and then with respect to GDP (in

U.S. dollars). The first normalized inflows proxies for the magnitude of the inclusion-driven inflow

shocks relative to the size of the sovereign debt market of each country. The hypothesis is that,

in countries whose sovereign debt markets are smaller, the impact of an inflow shock on sovereign

bond prices and yields could be larger than the impact of the same inflow shocks in countries with

larger sovereign debt markets. The second measure, that is, the estimated inflows relative to GDP,

measures the magnitude of the estimated inclusion-driven inflows relative to the overall size of

the country. We use this second measure to test for the hypothesis that countries experiencing a

larger sovereign debt inflow shock relative to the overall size of the country are the ones where the

domestic currency appreciates more.

Figure 10 presents the results from this analysis. Panel A shows the relationship, across coun-27In Table A3, Panel B, we present the results of a placebo test where we re-estimate Equation (3) using as dependentvariable the CARs in the interval [t − 3, t − 2], where t is the first trading day after the announcement. None of thecoefficients is statistically distinguishable from 0, thus further corroborating that the cross-country relationshipsestimated in Table 7 emerge only after the announcement episodes.

22

tries, between the 2-day change in the 5-year local currency sovereign bond yield after the announce-

ment and the estimated inflows relative to the size of the sovereign debt market. Panel B depicts

the relationship between the 2-day percentage appreciation of the domestic currency and the esti-

mated inflows relative to the GDP. In both cases, we see that sovereign bond yields and exchange

rates tend to respond more in countries experiencing larger relative inflow shocks. The only country

which appears to be in outlier in both figures is Czech Republic. However, this can be explained

by two things: first, Czech Republic is the only country that is being included in the GBI-EM, but

at the same time is being excluded from another index, the GBI-DM (for developed markets). J.P.

Morgan does not provide data on the amount of assets benchmarked against the GBI-DM, so we

cannot estimate the potential outflows due to this exclusion. As a result, the estimated net inflows

in Czech Republic are smaller than those we estimate; second, the 5-year sovereign bond yield is

Czech Republic is already close to the zero lower bound on the announcement date, which plausibly

reduces the scope for inflow shocks to reduce the yield.28

5.2 The Long-Term Consequences of Sovereign Debt Inflows

In this study, we assess the impact of sovereign debt inflow shocks on firms by looking at the

evolution of stock market prices in the days immediately following six major sovereign debt inflow

shocks in emerging countries. Our identification relies on the implicit assumption that stock market

prices efficiently respond to changes in the profitability and, more in general, in the future prospects

of firms operating in countries receiving such inflows. Yet, one might be interested in testing

whether the heterogeneous stock market effects documented in this paper effectively translate into

a heterogenous evolution of firms’ real variables in the long run. Addressing this issue requires

expanding the sample period much beyond the country inclusion announcement episodes, making

estimates more vulnerable to some identification concerns.28In fact one news article suggests that estimated inflows at the time of the announcement were between 3 and 6billions U.S. dollars. Our estimated inflows are 6.8 billion dollars, much larger than the average estimated netinflows. Source: Pio Online. https://www.pionline.com/article/20170428/ONLINE/170429837/j-p-morgan-drops-czech-republic-bonds-to-emerging-markets-indexes-inflows-expected (Retrieved on May 6, 2020).

23

Despite these potential concerns, in Pandolfi andWilliams (2020) we provide suggestive evidence

about the long-term effects of the country inclusion events, which are consistent with the results

of the present study. By looking at the evolution of domestic firms’ real variables in the years

around the inclusion episodes exploited in this study, in Pandolfi and Williams (2020) we show

that: i) financial and government-related firms grow significantly more, in terms of income and

number of employees, than firms operating in tradable industries in the 3 years following the

inclusion events; ii) financial and government-related firms tend to pay higher dividends than firms

in tradable industries after the inclusions; iii) more financially constrained firms also exhibit higher

profitability and pay more dividends in the post-inclusion period. The evidence in Pandolfi and

Williams (2020) therefore suggests that the heterogeneous stock market effects on domestic firms

of sovereign debt inflow shocks are the result of real changes in domestic firms’ future prospects.

6 A model of sovereign debt inflows

Our empirical results imply that sovereign debt inflows can have important distributional effects

in emerging economies, increasing the relative importance of the non-tradable sector vis-a-vis the

tradable sector. In this section, we present a model that formalizes this view.

6.1 The environment

We consider a small-open economy that lasts for two periods, t = 0, 1. The economy is populated

by a continuum of measure one of residents. All residents have identical preferences:

U(CN , CT ) =[ν · C

η−1η

N + (1− ν) · Cη−1η

T

] ηη−1

,

where CT and CN respectively denote consumption of tradable and non-tradable goods at t = 1,

and η is the elasticity of substitution between both goods. We assume throughout that η < 1.29

29See of Akinci (2017) for a review of the empirical evidence.

24

Production is organized in firms. In particular, we assume that there is a continuum of firms of

measure one in each sector, tradable and non-tradable. As we shall specify below, firms may need to

borrow funds in the credit market in order to produce. We assume that all credit is intermediated

by a continuum of competitive banks, with measure one. Firms and banks are all owned by the

economy’s residents.

Finally, there is a government that collects taxes from firms at t = 1 in order to repay pre-

existing public debt and provide subsidies.

6.1.1 Tradable and non-tradable production

Firms are indexed by sector T and N and by an individual firm index s ∈ [0, 1]. Each tradable

firm s is initially endowed with ωTs units of the tradable good. The total tradable endowment in

the economy is thus given by ωT =∫ 1

0 ωTs · ds.

At t = 1, the tradable good is produced competitively by final good producers that combine

capital and labor according to the following production function,

YT = Kα · L1−α,

where L denotes labor and

K =(∫ 1

0kε−1ε

s · ds) εε−1

is an aggregator of differentiated varieties of capital. Each variety of capital s is produced by the

respective tradable firm through investment. In particular, if tradable firm s invests is units of the

tradable good at t = 0 it produces

ks = A · is.

We think of ε as being high but finite.

Each resident is endowed with one unit of labor that is supplied inelastically in a competitive

25

labor market, so that L = 1.

In the non-tradable sector, firm s is endowed with ωNs units of the non-tradable good at t = 1.

The total non-tradable endowment is thus given by ωN =∫ 1

0 ωNs · ds.

6.1.2 Credit markets and the public sector

We index the banking sector with B and each individual bank with s ∈ [0, 1]. Banks start period

t = 0 with a pre-existing stock of public debt b as assets and deposits by residents d as liabilities.

A fraction λ of public debt is “long” and matures at t = 1; the remaining fraction 1− λ is “short”

and matures at t = 0. All deposits are “short” and mature at t = 0. We assume that d > (1− λ) ·b.

At t = 0, banks purchase public bonds, extend loans to tradable firms that wish to invest in

excess of their endowment, and repay maturing deposits. This is funded with deposits from other

tradable firms and from residents, with the proceeds from maturing short-term debt and through

the sale of long-term debt. Banks cannot borrow directly from the international financial market

(IFM) but they can sell public debt to the IFM. We assume, in particular, that the IFM spends

an exogenous amount x∗ in the secondary market for domestic public debt. The level of x∗ is key

and summarizes the magnitude of sovereign debt inflows in the model. We assume that x∗ ≤ ωT .

At t = 0, short-term public debt is rolled over at the market interest rate. At t = 1, the

government taxes firms and banks in order to pay all outstanding debt. In particular, we assume

that the profits of all firms and banks are taxed at an exogenous rate τ . Any fiscal surplus after

debt repayment is transferred back as a subsidy to a subset of “related” firms and banks. Since we

do not want government subsidies to affect the relative valuation of firms and banks, we assume

that the same fraction θ of firms in each sector and of banks are related to the government, and

whether a firm is related to the government or not is independent of its endowment.

26

Formally, the fiscal surplus after debt repayment is given by,

τ · (πN + πT + πB)− b · [λ+ (1− λ) ·R] ,

where πN , πT , and πB respectively denote profits of non-tradable firms, tradable firms, and banks.

This surplus is used to subsidize the subset of related firms and banks. For simplicity, we assume

that this subsidy is also proportional to profits, i.e., it can be interpreted as a reduction in the tax

rate. Letting σs denote the rate of the subsidy of a related firm or bank s, it follows that,

σs = 1θ·[τ − b · (λ+ (1− λ) ·R)

πN + πT + πB

],

which guarantees that the surplus is fully rebated back. To ensure that the government subsidy

does not affect the relative valuation of sectors, we assume that the subsidy is the same for all

related entities.

6.2 Equilibrium

To solve for equilibrium, we begin at t = 1 and then work our way backwards. At t = 1, the prices

of capital and labor are determined in competitive factor markets and equal:

w = (1− α) ·Kα and q = α ·Kα−1.

From profit maximization by the final good producers and market clearing, we can derive the

demand for each of the varieties of capital ks:

ps = q ·(K

ks

) 1ε

.

Pre-tax profits of tradable firms, non-tradable firms and banks are respectively given by:

πTs = ps · ks − (ks − ωTs) ·R,

27

πNs = pN · ωNs,

and

πBs = λ · b− [d− (1− λ) · b] ·R,

where R denotes the market interest rate.

At t = 0, tradable firms decide how much to invest as a function of the interest rate R. Taking

K as given, each tradable firm s ∈ [0, 1] maximizes its market value

VTs = πTsR· (1− τ + σs) =

[(A · psR− 1

)· is + ωs

]· (1− τ + σs) .

The optimal investment is given by

is =(ε− 1ε· α ·A

ε−1ε

R

)ε·K1+ε·(α−1).

It follows that is = i for all tradable firms s ∈ [0, 1], where i is also total investment. Taking

this into account, and noting that K = A · i, we can solve for

i =(ε− 1ε· α ·A

α

R

) 11−α

. (4)

Domestic residents roll over their deposits since they do not consume at t = 0. Market clearing

then requires that total investment equals firm endoments plus sovereign debt inflows x∗, i.e.

i = ωT + x∗. (5)

Equations (4) and (5) jointly determine the equilibrium interest rate:

R = ε− 1ε· α ·Aα · (ωT + x∗)α−1 .

28

Note that, as long as there is some degree of market power, i.e., ε <∞, the market interest rate is

lower than the marginal return of investment, i.e., α ·Aα · (ωT + x∗)α−1. The reason is that market

power depresses investment and thus the demand for credit.

Given the equilibrium levels of investment and interest rate, consumption of tradables and

non-tradables are respectively given by

CT = YT −R · x∗ = Aα · (ωT + x∗)α−1 ·(ωT +

(1− α · ε− 1

ε

)· x∗

), (6)

CN = ωN . (7)

Together with the relative demand of non-tradable goods

(CNCT

) 1η

= ν

1− ν ·1PN

,

Equations (6) and (7) allow us to derive the equilibrium price of non-tradables:

PN = ν

1− ν · ω− 1η

N ·[Aα · (ωT + x∗)α−1 ·

(ωT +

(1− α · ε− 1

ε

)· x∗

)] 1η

.

We can now compute the value of firms of different types. In the tradable sector, the value of

each firm s is given by

VTs =[ 1ε− 1 · (ωT + x∗) + ωTs

]· (1− τ + σs) .

The value of the average tradable firm equals

VT =∫ 1

0VTs · ds =

[ε

ε− 1 ·(ωT + x∗

ε

)]·[1− b · (λ+ (1− λ) ·R)

πN + πT + πB

].

29

In the non-tradable sector, the value of firm s is given by

VNs = PN · ωNsR

· (1− τ + σs) =

= κNs ·(Aα · (ωT + x∗)α−1

) 1−ηη ·

[(ωT +

(1− α · ε− 1

ε

)· x∗

)] 1η

· (1− τ + σs) ,

where κNs = ν1−ν ·

ε−1ε · ωNs · ω

− 1η

N . The value of the average non-tradable firm equals

VN =∫ 1

0VNs · ds =

= κN ·(Aα · (ωT + x∗)α−1

) 1−ηη ·

[(ωT +

(1− α · ε− 1

ε

)· x∗

)] 1η

·

·[1− b · (λ+ (1− λ) ·R)

πN + πT + πB

],

where κN = ν1−ν ·

ε−1ε · ω

1− 1η

N .

Finally, the equilibrium value of bank s is given by

VBs = b · λ− [d− b · (1− λ)] ·RR

· (1− τ + σs) ,

and the value of the average bank equals

VB =∫ 1

0VBs · ds =

=[

ε

ε− 1 ·b · λ · (ωT + x∗)1−α

α ·Aa− d+ (1− λ) · b

]·[1− b · (λ+ (1− λ) ·R)

πN + πT + πB

].

6.3 Winners and losers from inflows

We are interested in analyzing the impact of an increase in sovereign inflows x∗. The following

results follow from the model:

(i) In the tradable sector, the impact of sovereign inflows on firm values is highest for firms with

30

low endowment, i.e., 1VTs· dVTsdx∗

is decreasing in ωTs. In other words, the value of more financially

dependent firms (i.e. those that need to borrow more) increases more with x∗ than the value of

less financially dependent ones. The reason is that sovereign inflows reduce the interest rate and

thus the cost of borrowing.

(ii) On average, the value of non-tradable firms increases relatively more with x∗ than the value

of tradable firms, 1VN· dVNdx∗

>1VT· dVTdx∗

. This happens because sovereign inflows raise domestic

consumption of tradables and, thus, lead to a real appreciation.

(iii) The value of firms that are related to the government increases with x∗ more than the

value of firms that are unrelated to the government. In the model, this happens because sovereign

inflows raise the fiscal surplus, which is distributed among related firms. The fiscal surplus, in turn,

increases for two reasons. Sovereign inflows reduce the interest rate and allow the government to

roll-over its debt at a lower cost, which reduces government outlays. Moreover, capital inflows

are expansionary and this increases the tax base and thus government revenues. Ultimately, both

channels raise the government subsidy and hence the relative valuation of related to unrelated

firms.30

(iv) Finally, the valuation of banks increases with x∗ due to the capital gains on their holdings

of long-term government debt. Sovereign inflows reduce the interest rate and raise the price of

long-term bonds, which benefit banks. The higher the share of long-term bonds, the higher is the

increase in bank valuation relative to that firms. The higher is the leverage of banks, i.e., initial

deposits d relative to short-term debt, the higher is the increase in bank valuation relative to that

of firms.30Of course, our assumption that the government simply distributes its surplus as a handout among related firmsis an extreme simplification. We could think, at the cost of complicating the model, of other ways in which arelaxation of the government resource constraint would eventually benefit firms that are related to the government(i.e., through an expansion in the government’s demand for the goods produced by these firms).

31

7 Conclusions

This paper studies the effects of large sovereign debt inflows on domestic firms. To this end,

we exploit six announcements of country inclusion into two major local currency sovereign debt

market indexes. These announcements are not anticipated by investors and trigger large inflows

from international investors wishing to replicate the index compositions.

Our results show that sovereign debt inflows significantly reduce the local currency sovereign

bond yields and lead to an appreciation of the domestic currency. Also, they have sizable heteroge-

neous effects to domestic firms: financial and government-related firms exhibit larger CARs in the

2 days following country inclusion announcements; instead, firms operating in tradable industries

experience lower CARs. The former effect is more pronounced in countries where the 5-year gov-

ernment bond yields drops more after the event, whilst the latter effect is larger in countries where

the currency appreciates more.

Our findings shed novel light on the channels through which capital inflows to sovereign debt

markets affect firms in the economy. They highlight that sovereign debt inflows can have important

effects on the domestic economy of recipient countries. In addition, they can contribute to reshape

emerging economies, favoring the growth of the non-tradable sector at the expense of the tradable

one, promoting the development of the financial sector, and relaxing the financial constraints faced

by domestic firms.

32

References

Akinci, O., 2017. A Note on the Estimation of the Atemporal Elasticity of Substitution Between

Tradable and Nontradable Goods. mimeo.

Alfaro, L., Chari, A., Kanczuk, F., 2017. The real effects of capital controls: Firm-level evidence

from a policy experiment. Journal of International Economics 108, 191–210.

Alfaro, L., Kalemli-Ozcan, S., Volosovych, V., 2014. Sovereigns, Upstream Capital Flows, And

Global Imbalances. Journal of the European Economic Association 12, 1240–1284.

Altavilla, C., Pagano, M., Simonelli, S., 2017. Bank Exposures and Sovereign Stress Transmission.

Review of Finance 21, 2103–2139.

Andrade, S. C., Chhaochharia, V., 2018. The costs of sovereign default: Evidence from the stock

market. The Review of Financial Studies 31, 1707–1751.

Arellano, C., Bai, Y., Bocola, L., 2017. Sovereign Default Risk and Firm Heterogeneity. NBER

Working Papers 23314, National Bureau of Economic Research, Inc.

Arslanalp, S., Tsuda, T., 2014. Tracking Global Demand for Emerging Market Sovereign Debt.

IMF Working Papers 14/39, International Monetary Fund.

Bank for International Settlements (BIS), 2019. Monetary Policy Frameworks in EMES: Inflation

Targeting, the Exchange Rate and Financial Stability. In: Annual Economic Report.

Basak, S., Pavlova, A., 2013. Asset Prices and Institutional Investors. American Economic Review

103, 1728–1758.

Baskaya, Y. S., di Giovanni, J., Kalemli-Ozcan, S., Peydro, J.-L., Ulu, M. F., 2017. Capital flows

and the international credit channel. Journal of International Economics 108, 15–22.

Becker, B., Ivashina, V., 2018. Financial repression in the european sovereign debt crisis. Review

of Finance 22, 83–115.

33

Benigno, G., Converse, N., Fornaro, L., 2015. Large capital inflows, sectoral allocation, and eco-

nomic performance. Journal of International Money and Finance 55, 60–87.

Blanchard, O., Ostry, J. D., Ghosh, A. R., Chamon, M., 2017. Are Capital Inflows Expansionary

or Contractionary? Theory, Policy Implications, and Some Evidence. IMF Economic Review 65,

563–585.

Borri, N., Shakhnov, K., 2018. Limited Participation and Local Currency Sovereign Debt. Unpub-

lished, available at SSRN: https://ssrn.com/abstract=2978127.

Bottero, M., Lenzu, S., Mezzanotti, F., 2020. Sovereign debt exposure and the bank lending channel:

impact on credit supply and the real economy. Journal of International Economics, forthcoming.

Broner, F., Clancy, D., Erce, A., Martin, A., 2019. Fiscal Multipliers and Foreign Holdings of Public

Debt. Working Paper Series 2255, European Central Bank.

Broner, F., Erce, A., Martin, A., Ventura, J., 2014. Sovereign Debt Markets in Turbulent Times:

Creditor Discrimination and Crowding-out Effects. Journal of Monetary Economics 61, 114–142.

Cakici, N., Fabozzi, F. J., Tan, S., 2013. Size, value, and momentum in emerging market stock

returns. Emerging Markets Review 16, 46–65.

Calomiris, C. W., Larrain, M., Schmukler, S. L., 2020. Capital Inflows, Equity Issuance Activity,

and Corporate Investment. Journal of Financial Intermediation, forthcoming.

Chari, A., Blair Henry, P., 2008. Firm-specific information and the efficiency of investment. Journal

of Financial Economics 87, 636–655.

Chari, A., Henry, P. B., 2004. Risk Sharing and Asset Prices: Evidence from a Natural Experiment.

Journal of Finance 59, 1295–1324.

Chari, A., Leary, R., Phan, T., 2018. The Costs of (sub)Sovereign Default Risk: Evidence from

Puerto Rico. Working Paper 18-3, Federal Reserve Bank of Richmond.

Cremers, M., Ferreira, M. A., Matos, P., Starks, L., 2016. Indexing and Active Fund Management:

International Evidence. Journal of Financial Economics 120, 539–560.

34

Dittmar, R. F., 2008. Do Sovereign Bonds Benefit Corporate Bonds in Emerging Markets? Review

of Financial Studies 21, 1983–2014.

Du, W., Schreger, J., 2016. Local Currency Sovereign Risk. Journal of Finance 71, 1027–1070.

Gabaix, X., Maggiori, M., 2015. International Liquidity and Exchange Rate Dynamics. The Quar-

terly Journal of Economics 130, 1369–1420.

Gennaioli, N., Martin, A., Rossi, S., 2014. Sovereign Default, Domestic Banks, and Financial

Institutions. Journal of Finance 69, 819–866.

Gopinath, G., Boz, E., Casas, C., Díez, F. J., Gourinchas, P.-O., Plagborg-Moller, M., 2020.

Dominant Currency Paradigm. American Economic Review 110, 677–719.

Greenwood, R., Vayanos, D., 2010. Price Pressure in the Government Bond Market. American

Economic Review 100, 585–590.

Hébert, B., Schreger, J., 2017. The Costs of Sovereign Default: Evidence from Argentina. American

Economic Review 107, 3119–3145.

Hofmann, B., Shim, I., Shin, H. S., 2019. Bond risk premia and the exchange rate. BIS Working

Papers 775, Bank for International Settlements.

Kashyap, A. K., Kovrijnykh, N., Li, J., Pavlova, A., 2018. The Benchmark Inclusion Subsidy.

NBER Working Papers 25337, National Bureau of Economic Research, Inc.

Kohn, D., 2015. Addicted to Debt: Foreign Purchases of U.S. Treasuries and the Term-Premium.

Department of Economics Working Papers 2015-1, Universidad Torcuato Di Tella.

Kolasa, M., Wesolowski, G., 2020. International spillovers of quantitative easing. Journal of Inter-

national Economics, forthcoming.

Krishnamurthy, A., Nagel, S., Vissing-Jorgensen, A., 2017. ECB Policies Involving Government

Bond Purchases: Impact and Channels. NBER Working Papers 23985, National Bureau of Eco-

nomic Research, Inc.

35

Krishnamurthy, A., Vissing-Jorgensen, A., 2011. The Effects of Quantitative Easing on Interest

Rates: Channels and Implications for Policy. Brookings Papers on Economic Activity 42, 215–

287.

Lane, P. R., McQuade, P., 2014. Domestic credit growth and international capital flows. The

Scandinavian Journal of Economics 116, 218–252.

Larrain, M., 2015. Capital Account Opening and Wage Inequality. Review of Financial Studies 28,

1555–1587.

Mian, A., Sufi, A., 2014. What Explains the 2007-2009 Drop in Employment? Econometrica 82,

2197–2223.

Morelli, J., Perez, D., Ottonello, P., 2019. Global Banks and Systemic Debt Crises. 2019 Meeting

Papers 644, Society for Economic Dynamics.

Obstfeld, M., Rogoff, K., 1995. Exchange Rate Dynamics Redux. Journal of Political Economy 103,

624–660.

Ongena, S., Popov, A., Horen, N. V., 2019. The Invisible Hand of the Government: Moral Suasion

during the European Sovereign Debt Crisis. American Economic Journal: Macroeconomics 11,

346–379.

Pandolfi, L., Williams, T., 2019. Capital Flows and Sovereign Debt Markets: Evidence from Index

Rebalancings. Journal of Financial Economics 132, 384–403.

Pandolfi, L., Williams, T., 2020. Real Effects of Sovereign Debt Inflows. AEA Papers and Proceed-

ings, forthcoming.

Priftis, R., Zimic, S., 2020. Sources of Borrowing and Fiscal Multipliers. The Economic Journal

Ueaa051.

Raddatz, C., Schmukler, S., Williams, T., 2017. International Asset Allocations and Capital Flows:

The Benchmark Effect. Journal of International Economics 108, 413–430.

36

Rajan, R. G., Zingales, L., 1998. Financial Dependence and Growth. American Economic Review

88, 559–586.

Schnabl, P., 2012. The International Transmission of Bank Liquidity Shocks: Evidence from an

Emerging Market. The Journal of Finance 67, 897–932.

Vayanos, D., Vila, J.-L., 2009. A Preferred-Habitat Model of the Term Structure of Interest Rates.

CEPR Discussion Papers 7547, C.E.P.R. Discussion Papers.

Warnock, F., Warnock, V., 2009. International Capital Flows and U.S. Interest Rates. Journal of

International Money and Finance 28, 903–919.

Williams, T., 2018. Capital inflows, sovereign debt and bank lending: Micro-evidence from an

emerging market. Review of Financial Studies 31, 4958–4994.

37

Figure 1: Stock Prices around Events

99

100

101

102

103

Pri

ce (

nor

mal

ized

)

-7 0 7 14 21

Days to/from Announcement Date

Financial Government-related Tradable

Note: This figure depicts the evolution of the stock prices of financial firms, government-related firms, and firms operating intradable sectors, separately, over a time window which starts 7 days before the announcement of each country’s inclusion inthe corresponding index, and ends 14 days after it. Stock prices are normalized to their values in the last trading day beforeeach announcement episode, indicated by a vertical dashed line. Observations in the top and the bottom percentile of thecountry-by-date distribution of stock prices are excluded.

38

Figure 2: Foreign Share of Sovereign Debt around Events

15%

20%

25%

30%

-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6

Quarters to/from Announcement Quarter

Observed values Pre-announcement trend

Average Private Foreign Share of Government Debt

15%

20%

25%

30%

-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6

Quarters to/from Announcement Quarter

Observed values Pre-announcement trend

Median Private Foreign Share of Government Debt

This figure depicts the quarterly time series of the average and the median (left panel and right panel, respectively) share of sovereign debt held by foreign privateinvestors for Colombia, Mexico, Romania and South Africa over a time window which starts 6 quarters before the announcement of each country’s inclusion in thecorresponding index, and ends 6 quarters after it. In both panels, the dotted line is a linear trend estimated on the pre-announcement period. The vertical dashed lineindicates the quarter prior to the announcement episodes. Data is from Arslanalp and Tsuda (2014).

39

Figure 3: Balance of Payments: Private vs. Public Inflows

0

1%

2%

3%

4%

Inflow

s over

GD

P

Private Inflows Public Inflows

Average in the 3 yearsbefore the announcement

Year of theannouncement

This figure depicts the average private and public net inflows to Colombia, Czech Republic, Mexico, Nigeria, Romania andSouth Africa in the year of the announcement of each country’s inclusion into the corresponding index vs. the average of publicand private inflows to these countries in the three years before the announcement episodes. Private inflows are the sum of foreigndirect investments, portfolio equity net inflows and private debt net inflows. Public inflows are net inflows to the countries’sovereign debt markets. Both are in U.S. dollars and are normalized by the GDP of each country, before being averaged acrosscountries. Data is from the IMF Balance of Payments Statistics and IMF WEO.

40

Figure 4: 5-Year Government Bond Yields around Events

5.5

6

6.5

7

5Y G

over

nm

ent

Bond Y

ield

(%

)

27jan2014 21feb2014 18mar2014 12apr2014 07may2014

Colombia

-.4

-.2

0

.2

5Y G

over

nm

ent

Bond Y

ield

(%

)

03jan2017 28jan2017 22feb2017 19mar2017 13apr2017

Czech Republic

6.4

6.6

6.8

7

7.2

5Y G

over

nm

ent

Bond Y

ield

(%

)

08feb2010 05mar2010 30mar2010 24apr2010 19may2010

Mexico

12

13

14

15

16

17

5Y G

over

nm

ent

Bon

d Y

ield

(%

)

26jun2012 21jul2012 15aug2012 09sep2012 04oct2012

Nigeria

5.5

6

6.5

7

5Y G

over

nm

ent

Bon

d Y

ield

(%

)

26nov2012 21dec2012 15jan2013 09feb2013 06mar2013

Romania

6.8

7

7.2

7.4

7.6

5Y G

over

nm

ent

Bon

d Y

ield

(%

)

26feb2012 22mar2012 16apr2012 11may2012 05jun2012

South Africa

This figure depicts the time series of the 5-year local currency government bond yield in Colombia, Czech Republic, Mexico, Nigeria, Romania and South Africa over atime window which starts 2 months before the announcement of each country’s inclusion in the corresponding index, and ends 2 months after it. The vertical dashedline indicates the last trading day before each announcement episode.

41

Figure 5: Exchange Rates around Events

94

96

98

100

102

Exch

ange

Rat

e (n

orm

aliz

ed)

27jan2014 21feb2014 18mar2014 12apr2014 07may2014

Colombia

97

98

99

100

101

102

Exch

ange

Rat

e (n

orm

aliz

ed)

03jan2017 28jan2017 22feb2017 19mar2017 13apr2017

Czech Republic

98

100

102

104

106

Exch

ange

Rat

e (n

orm

aliz

ed)

08feb2010 05mar2010 30mar2010 24apr2010 19may2010

Mexico

100

101

102

103

104

Exch

ange

Rat

e (n

orm

aliz

ed)

26jun2012 21jul2012 15aug2012 09sep2012 04oct2012

Nigeria

98

100

102

104

106

108

Exch

ange

Rat

e (n

orm

aliz

ed)

26nov2012 21dec2012 15jan2013 09feb2013 06mar2013

Romania

90

95

100

105

110

Exch

ange

Rat

e (n

orm

aliz

ed)

26feb2012 22mar2012 16apr2012 11may2012 05jun2012

South Africa

This figure depicts the time series of the exchange rate in Colombia, Czech Republic, Mexico, Nigeria, Romania and South Africa over a time window which starts 2months before the announcement of each country’s inclusion in the corresponding index, and ends 2 months after it. The exchange rate for each country is the amountof local currency needed to buy 1 U.S. dollar, and is normalized to its value in the last trading day before each announcement episode, indicated by a vertical dashedline.

42

Figure 6: Distribution of 2-day Changes in 5Y Government Bond Yields

0

20

40

60

80

100

Fre

quen

cy (

%)

-50 -40 -30 -20 -10 0 10 20 30 40 50

2-Day Change 5Y Government Bond Yield (basis points)

Colombia

0

20

40

60

80

100

Fre

quen

cy (

%)

-20 -10 0 10 20

2-Day Change 5Y Government Bond Yield (basis points)

Czech Republic

0

20

40

60

80

Fre

quen

cy (

%)

-30 -20 -10 0 10 20 30

2-Day Change 5Y Government Bond Yield (basis points)

Mexico

0

20

40

60

80

Fre

quen

cy (

%)

-100 -80 -60 -40 -20 0 20 40 60 80 100

2-Day Change 5Y Government Bond Yield (basis points)

Nigeria

0

50

100

150

Fre

quen

cy (

%)

-100 -80 -60 -40 -20 0 20 40 60 80 100

2-Day Change 5Y Government Bond Yield (basis points)

Romania

0

20

40

60

80

Fre

quen

cy (

%)

-40 -30 -20 -10 0 10 20 30 40

2-Day Change 5Y Government Bond Yield (basis points)

South Africa

This figure depicts the distribution of 2-day changes in the 5-year local currency government bond yield in Colombia, Czech Republic, Mexico, Nigeria, Romania andSouth Africa in the 2 years around the announcement of each country’s inclusion in the corresponding index. The vertical line in each panel indicates the change in the5-year government bond yield in the 2 days following the announcement episode in each country.

43

Figure 7: Distribution of 2-day Percentage Changes in Exchange Rates

0

20

40

60

80

Fre

quen

cy (

%)

-5 -4 -3 -2 -1 0 1 2 3 4 5

2-Day Percentage Change in Exchange Rate (percentage points)

Colombia

0

20

40

60

80

Fre

quen

cy (

%)

-4 -3 -2 -1 0 1 2 3 4

2-Day Percentage Change in Exchange Rate (percentage points)

Czech Republic

0

20

40

60

80

Fre

quen

cy (

%)

-5 -4 -3 -2 -1 0 1 2 3 4 5

2-Day Percentage Change in Exchange Rate (percentage points)

Mexico

0

50

100

150

200

Fre

quen

cy (

%)

-4 -3 -2 -1 0 1 2 3 4

2-Day Percentage Change in Exchange Rate (percentage points)

Nigeria

0

20

40

60

80

Fre

quen

cy (

%)

-4 -3 -2 -1 0 1 2 3 4

2-Day Percentage Change in Exchange Rate (percentage points)

Romania

0

10

20

30

40

50

Fre

quen

cy (

%)

-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6

2-Day Percentage Change in Exchange Rate (percentage points)

South Africa

This figure depicts the distribution of 2-day changes in the log of the exchange rate in Colombia, Czech Republic, Mexico, Nigeria, Romania and South Africa in the 2years around the announcement of each country’s inclusion in the corresponding index. The vertical line in each panel indicates the percentage change in the exchangerate in the 2 days following the announcement episode in each country.

44

Figure 8: Cumulative Abnormal Returns around Events

-0.5%

0

0.5%

1%

Cum

ula

tive

Abnor

mal

Ret

urn

s

-3 -2 -1 0 1 2 3 4 5

Days to/from Announcement Date

1-Factor Model 2-Factor Model

5-Factor Model

All firms

-0.5%

0

0.5%

1%

Cum

ula

tive

Abnor

mal

Ret

urn

s

-3 -2 -1 0 1 2 3 4 5

Days to/from Announcement Date

Gvt&Fin (1-Factor Model) Gvt&Fin (2-Factor Model) Gvt&Fin (5-Factor Model)

Trad (1-Factor Model) Trad (2-Factor Model) Trad (5-Factor Model)

Financial & Government-related vs. tradable firms

This figure depicts the evolution of the cumulative abnormal returns of domestic firms over a time window which starts 3 days before the announcement of each country’sinclusion in the corresponding index, and ends 5 days after it. The left panel depicts the cumulative abnormal returns of all firms, while the right panel depicts thecumulative abnormal returns of financial and government-related firms, and tradable firms, separately. Cumulative abnormal returns are computed using three models.The only risk factor in the 1-factor model is the return of the MSCI Emerging Markets Index; the two risk factors in the 2-factor model are the return of the MSCIEmerging Markets Index and the return of the MSCI World Index; the 5-factor models include these two risk factors plus small minus big (SMB), high minus low(HML), and momentum (WML) factors. The vertical dashed line indicates the last trading day before the announcement. Observations in the top and the bottompercentile of the country-by-date distribution of cumulative abnormal returns are excluded.

45

Figure 9: Heterogeneous Effects of the Shocks

CO

CZ

MX

NG

RO

ZA

.2

.4

.6

.8

1

1.2

β1 (G

ovt&

Fin

)

-80 -60 -40 -20 0

2-Day Change 5Y-Government Bond Yield (basis points)

CO

CZ

MX

NGRO

ZA-1

-.5

0

.5

β2 (T

radab

le)

-2 -1.5 -1 -.5 0

2-Day Percentage Change in Exchange Rate (percentage points)

This figure depicts the relationship between the size of the 2-day changes in sovereign bond yields and exchange rates (in the two days following the announcementof each country’s inclusion in the corresponding index), and the stock market effects on government-related and financial firms (left panel), and tradable firms (rightpanel), separately. β1 and β2 are the OLS estimated coefficients of 6 country-specific regressions of the form: CAR = β11(Govt&Fin) + β21(Tradable) + β3EFD + ε,where CAR is the cumulative abnormal returns of each firm in the two days following the announcement episode (computed using a 1-factor model where the onlyrisk factor is the return of the MSCI Emerging Markets Index), 1(Govt&Fin) is a dummy variable which equals 1 for government-related and financial companies,1(Tradable) is a dummy which equals 1 for firms operating in tradable sectors, and EFD is a measure of external financial dependance computed at the industry-level.Observations in the top and the bottom percentile of the country-specific distribution of 2-day cumulative abnormal returns are excluded.

46

Figure 10: Estimated Inflows and Size of the Shocks

CO

CZMX

NG

RO

ZA

-100

-80

-60

-40

-20

0

2-D

ay C

han

ge 5

Y-G

over

nm

ent

Bon

d Y

ield

(bps)

10% 15% 20% 25% 30%

Estimated Inflows over Market Value of Local-currency Sov. Debt

All countries Excluding Czech Republic

CO

CZ

MX

NG

RO

ZA

-2.2

-1.8

-1.4

-1

-.6

-.2

2-D

ay L

og C

han

ge F

X (

per

centa

ge p

oints

)

0 0.5% 1% 1.5% 2% 2.5% 3% 3.5%

Estimated Inflows over GDP

All countries Excluding Czech Republic

This figure depicts the relationship between the size of the 2-day changes in sovereign bond yields and exchange rates (in the two days following the announcement ofeach country’s inclusion in the corresponding index), and the estimated sovereign debt inflows relative to the size of each country’s local-currency sovereign debt market(left panel) and each country’s GDP (right panel), separately. The estimated inflows, the market value of sovereign debt markets, and countries’ GDP are in US dollars.Estimated inflows are computed as the change in benchmark weight following the inclusion of each country in the index, calculated over the entire implementationperiod, multiplied by the assets under management of funds tracking their returns against the corresponding index. The dashed and the solid line are regression linesdescribing the relationship between estimated inflows and shocks, with and without Czech Republic, respectively.

47

Table 1: Summary Statistics

Mean St. Dev. CO CZ MX NG RO ZA

Financial 0.14 0.35 0.36 0.21 0.14 0.17 0.10 0.10

Government-Related 0.09 0.29 0.19 0.36 0.07 0.11 0.05 0.08

Tradable 0.35 0.48 0.19 0.21 0.23 0.42 0.54 0.32

External Financial Dependance -2.04 2.91 -2.53 -1.88 -2.04 -2.19 -1.65 -2.05

Foreign Ownership 0.11 0.31 0.18 0.21 0.16 0.09 0.02 0.12

Issue Debt 0.16 0.36 0.38 0.36 0.47 0.04 0.05 0.09

Log(Assets) 15.44 3.25 21.68 16.57 16.43 17.41 12.18 14.33

Observations 909 72 14 137 170 155 361

This table reports summary statistics about domestic firms in our sample. The first two columns report themean and the standard deviation of our main explanatory variables computed over the entire sample of firms.Columns 3 to 8 report the average of these variables in each of the countries in our sample (Colombia, CzechRepublic, Mexico, Nigeria, Romania, and South Africa). Financial is an indicator variable that is equal to 1for financial firms. Government-Related is an indicator variable that is equal to 1 for government-related firms.Tradable is an indicator variable that is equal to 1 for firms in tradable industries (according to the classificationin Mian and Sufi (2014)). External Financial Dependance is a measure of firms’ dependance on external financingsources, computed following Rajan and Zingales (1998). Foreign Ownership is an indicator variable that is equalto 1 for firms which are included in the MSCI Emerging Markets Index in the year prior to the announcementdate. Issue Debt is an indicator variable that is equal to 1 for firms that issued corporate debt or obtained asyndicated loan in the years before the announcement date.

48

Table 2: Shocks

Colombia Czech Rep. Mexico Nigeria Romania South Africa

∆Yield (bps) -30.190 -7.300 -4.800 -42.000 -86.900 -23.600

∆Yield−∆Yield -30.649*** -7.641*** -4.569*** -42.870*** -85.819*** -22.674***

%∆ExchRate (pp) -2.026 -0.532 -1.266 -0.057 -0.633 -2.006

%∆ExchRate−%∆ExchRate -2.179*** -0.484*** -1.200*** -0.077*** -0.635*** -2.120***

This table reports the changes in government bond yields and exchange rates in the 2 days following the announcement of eachcountry’s inclusion in the corresponding index. ∆Y ield(bps) is the 2-day change in the 5-year local currency government bondyield in basis points. %∆ExchRate(pp) is the 2-day percentage change in the exchange rate (computed as the difference in the logof the exchange rate) in percentage points. ∆Y ield−∆Y ield and %∆ExchRate−%∆ExchRate are the differences between thechanges in government bond yields and exchange rates in the 2 days following the announcement episodes, and the average 2-daychanges of these two variables in the 2 years around the announcement of each country’s inclusion in the corresponding index. *,**, ***, denote that these differences are statistically different from 0 at the 10%, 5%, and 1% confidence level, respectively.

49

Table 3: Aggregate Cumulative Abnormal Returns

Panel A: Cumulative Abnormal Returns after Announcement Dates

Demeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

Average 2-day CAR 0.219** 0.133 0.111 0.117

(0.099) (0.102) (0.102) (0.108)

Number of Countries 6 6 6 6

Observations 861 861 861 861

Panel B: Cumulative Abnormal Returns before Announcement Dates

Demeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

Average 2-day CAR 0.117 0.014 0.041 -0.006

(0.096) (0.096) (0.096) (0.097)

Number of Countries 6 6 6 6

Observations 861 861 861 861

This table reports the average cumulative abnormal return of firms in our sample after and before the announce-ment of each country’s inclusion in the corresponding index (Panel A and Panel B, respectively). The averageCAR in Panel A is computed as the average CAR in the 2 days following the announcement. The average CARin Panel B is computed as the average CAR over the interval [t − 3, t − 2], where t is the first trading day afterthe announcement. CARs in the first column (the demeaned returns) are computed as the cumulated differencesbetween the daily returns and the average daily return in the year preceding the announcement. CARs in columns2 to 4 are computed using three different factor models. The only risk factor in the 1-factor model is the returnof the MSCI Emerging Markets Index; the two risk factors in the 2-factor model are the return of the MSCIEmerging Markets Index and the return of the MSCI World Index; the 5-factor model include these two riskfactors plus small minus big (SMB), high minus low (HML), and momentum (WML) factors. Observations inthe top and the bottom percentile of the country-specific distribution of 2-day cumulative abnormal returns areexcluded. Standard errors in parenthesis are clustered at the country-by-industry level. *, **, ***, denote thatthe the average CAR is statistically different from 0 at the 10%, 5%, and 1% confidence level, respectively.

50

Table 4: Cumulative Abnormal Returns, by Firm Type

Panel A: Financial firms

Demeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

Average 2-day CAR 0.688*** 0.666*** 0.643*** 0.607***

(0.181) (0.173) (0.166) (0.183)

Number of Countries 6 6 6 6

Observations 118 118 118 118

Panel B: Government-related firms

Demeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

Average 2-day CAR 0.909*** 0.802*** 0.831*** 0.866***

(0.237) (0.245) (0.247) (0.250)

Number of Countries 6 6 6 6

Observations 78 78 78 78

Panel C : Tradable firms

Demeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

Average 2-day CAR -0.132 -0.205 -0.239 -0.202

(0.173) (0.166) (0.165) (0.177)

Number of Countries 6 6 6 6

Observations 306 305 305 304

This table reports the average cumulative abnormal return of financial firms (Panel A), government-related firms(Panel B), and firms operating in tradable sectors (Panel C) in the 2 days following the announcement of eachcountry’s inclusion in the corresponding index. CARs in the first column (the demeaned returns) are computedas the cumulated differences between the daily returns and the average daily return in the year preceding theannouncement. CARs in columns 2 to 4 are computed using three different factor models. The only risk factorin the 1-factor model is the return of the MSCI Emerging Markets Index; the two risk factors in the 2-factormodel are the return of the MSCI Emerging Markets Index and the return of the MSCI World Index; the 5-factormodel include these two risk factors plus small minus big (SMB), high minus low (HML), and momentum (WML)factors. Observations in the top and the bottom percentile of the country-specific distribution of 2-day cumulativeabnormal returns are excluded. Standard errors in parenthesis are clustered at the country-by-industry level. *,**, ***, denote that the the average CAR is statistically different from 0 at the 10%, 5%, and 1% confidence level,respectively.

51

Table 5: Main Results

2-Day Cumulative Abnormal Returns after Announcement Dates

Demeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

Financial 0.525* 0.613** 0.614** 0.573*

(0.290) (0.285) (0.280) (0.302)

Government-Related 0.632*** 0.598** 0.656*** 0.700***

(0.241) (0.249) (0.247) (0.252)

Tradable -0.514** -0.472** -0.495** -0.450*

(0.218) (0.220) (0.220) (0.235)

External Financial Dependance 0.072* 0.070* 0.075* 0.073*

(0.041) (0.041) (0.040) (0.043)

Financial – Tradable 1.04*** 1.09*** 1.11*** 1.02***

(0.31) (0.30) (0.29) (0.31)

Govt-Related – Tradable 1.15*** 1.07*** 1.15*** 1.15***

(0.32) (0.33) (0.32) (0.33)

Number of Countries 6 6 6 6

Observations 857 857 857 857

R2 0.02 0.02 0.02 0.02

This table reports the OLS coefficients of a regression where the dependent variable is the CAR of each firm in the 2 daysfollowing the announcement of each country’s inclusion in the corresponding index. The explanatory variables are: Financial,which is an indicator variable that is equal to 1 for financial firms; Government-Related, which is an indicator variable thatis equal to 1 for government-related firms; Tradable, which is an indicator variable equal to for firms in tradable industries(according to the classification in Mian and Sufi (2014)); and External Financial Dependance, which is a measure of firms’dependance on external financing sources, computed following Rajan and Zingales (1998). The table reports also the differencebetween the estimated coefficient on Financial and that on Tradable, as well as the difference between the estimated coefficientson Government-Related and that on Tradable. CARs in the first column (the demeaned returns) are computed as the thecumulated differences between the daily returns and the average daily return in the year preceding the announcement. CARsin columns 2 to 4 are computed using three different factor models. The only risk factor in the 1-factor model is the return ofthe MSCI Emerging Markets Index; the two risk factors in the 2-factor model are the return of the MSCI Emerging MarketsIndex and the return of the MSCI World Index; the 5-factor model include these two risk factors plus small minus big (SMB),high minus low (HML), and momentum (WML) factors. Observations in the top and the bottom percentile of the country-specific distribution of 2-day cumulative abnormal returns are excluded. Standard errors in parenthesis are clustered at thecountry-by-industry level. *, **, ***, denote significance at the 10%, 5%, and 1% confidence level, respectively.

52

Table 6: Additional Controls

2-Day Cumulative Abnormal Returns after Announcement DatesDemeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

Financial 0.497* 0.494 0.584** 0.618* 0.569** 0.557* 0.578* 0.539(0.276) (0.339) (0.275) (0.336) (0.271) (0.329) (0.297) (0.364)

Government-Related 0.505** 0.451* 0.500** 0.441 0.569** 0.498* 0.641*** 0.557**(0.237) (0.264) (0.244) (0.270) (0.243) (0.271) (0.245) (0.271)

Tradable -0.431* -0.539** -0.398* -0.499* -0.421* -0.533* -0.364 -0.431(0.226) (0.273) (0.226) (0.279) (0.227) (0.279) (0.246) (0.307)

External Financial Dependance 0.074* 0.083* 0.072* 0.083* 0.076* 0.087* 0.076* 0.085*(0.041) (0.048) (0.042) (0.050) (0.041) (0.049) (0.043) (0.051)

Foreign Ownership 0.222 0.112 0.073 0.002 0.049 -0.092 0.056 -0.089(0.243) (0.262) (0.255) (0.278) (0.249) (0.277) (0.249) (0.283)

Issue Debt 0.133 0.087 0.040 0.010 0.008 -0.049 -0.042 -0.113(0.246) (0.266) (0.253) (0.274) (0.268) (0.295) (0.279) (0.314)

Log(Assets) 0.026 0.013 0.039 0.046(0.046) (0.046) (0.049) (0.060)

Country FE Yes Yes Yes Yes Yes Yes Yes YesFinancial – Tradable 0.93*** 1.03*** 0.98*** 1.12*** 0.99*** 1.09*** 0.94*** 0.97**

(0.29) (0.37) (0.29) (0.36) (0.28) (0.35) (0.31) (0.39)Govt-Related – Tradable 0.94*** 0.99*** 0.90*** 0.94** 0.99*** 1.03*** 1.01*** 0.99**

(0.32) (0.37) (0.33) (0.39) (0.33) (0.39) (0.34) (0.40)Number of Countries 6 6 6 6 6 6 6 6Observations 857 663 857 663 857 663 857 664R2 0.04 0.04 0.04 0.04 0.04 0.04 0.03 0.03This table reports the OLS coefficients of a regression where the dependent variable is the CAR of each firm in the 2days following the announcement of each country’s inclusion in the corresponding index. The explanatory variables are:Financial, which is an indicator variable that is equal to 1 for financial firms; Government-Related, which is an indicatorvariable that is equal to 1 for government-related firms; Tradable, which is an indicator variable equal to for firms intradable industries (according to the classification in Mian and Sufi (2014)); External Financial Dependance, which isa measure of firms’ dependance on external financing sources, computed following Rajan and Zingales (1998); ForeignOwnership, which is an indicator variable that is equal to 1 for firms which are included in the MSCI Emerging MarketsIndex in the year prior to the announcement date; IssueDebt, which is an indicator variable that is equal to 1 for firmsthat issued corporate debt or obtained a syndicated loan in the years before the announcement date; and Log(Assets),which is the logarithm of the total value of a firm’s assets. All regressions include country fixed effects. The table reportsalso the difference between the estimated coefficient on Financial and that on Tradable, as well as the difference betweenthe estimated coefficients on Government-Related and that on Tradable. CARs in the first column (the demeaned returns)are computed as the cumulated differences between the daily returns and the average daily return in the year precedingthe announcement. CARs in columns 2 to 4 are computed using three different factor models. The only risk factor inthe 1-factor model is the return of the MSCI Emerging Markets Index; the two risk factors in the 2-factor model are thereturn of the MSCI Emerging Markets Index and the return of the MSCI World Index; the 5-factor model include thesetwo risk factors plus small minus big (SMB), high minus low (HML), and momentum (WML) factors. Observations inthe top and the bottom percentile of the country-specific distribution of 2-day cumulative abnormal returns are excluded.Standard errors in parenthesis are clustered at the country-by-industry level. *, **, ***, denote significance at the 10%,5%, and 1% confidence level, respectively.

53

Table 7: Interaction with Size of the Shocks

2-Day Cumulative Abnormal Returns after Announcement Dates

Demeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

∆Yield×Govt&Fin 0.015*** 0.017*** 0.015*** 0.017*** 0.015*** 0.017*** 0.015*** 0.017***

(0.005) (0.006) (0.005) (0.006) (0.005) (0.006) (0.005) (0.006)

%∆ExchRate× Tradable -0.354** -0.411*** -0.341** -0.393*** -0.358** -0.414*** -0.295* -0.349**

(0.147) (0.144) (0.145) (0.144) (0.146) (0.144) (0.166) (0.162)

External Financial Dependance 0.070* 0.066 0.070* 0.068*

(0.038) (0.040) (0.039) (0.041)

Country FE Yes Yes Yes Yes Yes Yes Yes Yes

Number of Countries 6 6 6 6 6 6 6 6

Observations 861 857 861 857 861 857 861 857

R2 0.04 0.04 0.04 0.04 0.04 0.04 0.03 0.04

This table reports the OLS coefficients of a regression where the dependent variable is the CAR of each firm in the 2 days followingthe announcement of each country’s inclusion in the corresponding index. The explanatory variables are: ∆Y ield × Govt&Fin,which is the product of indicator variable that is equal to 1 for financial and government-related firms, and the 2-day change inthe 5-year local currency government bond yield, in basis points; %∆ExchRate × Tradable, which is the product of indicatorvariable that is equal to 1 for firms in tradable industries (according to the classification in Mian and Sufi (2014)), and the 2-daypercentage change in the exchange rate (computed as the difference in the log of the exchange rate) in percentage points; andExternal Financial Dependance, which is a measure of firms’ dependance on external financing sources, computed following Rajanand Zingales (1998). All regressions include country fixed effects. CARs in the first column (the demeaned returns) are computedas the cumulated differences between the daily returns and the average daily return in the year preceding the announcement. CARsin columns 2 to 4 are computed using three different factor models. The only risk factor in the 1-factor model is the return of theMSCI Emerging Markets Index; the two risk factors in the 2-factor model are the return of the MSCI Emerging Markets Index andthe return of the MSCI World Index; the 5-factor models include these two risk factors plus small minus big (SMB), high minuslow (HML), and momentum (WML) factors. Observations in the top and the bottom percentile of the country-specific distributionof 2-day cumulative abnormal returns are excluded. Standard errors in parenthesis are clustered at the country-by-industry level.*, **, ***, denote significance at the 10%, 5%, and 1% confidence level, respectively.

54

A Appendix

A.1 International Benchmark Indexes for Local Currency Sovereign Bond Mar-

kets

International indexes are indexes which combine and track assets of different classes from different

countries. Depending on the criteria used to select the countries to be included in the index,

international indexes can be categorized in regional and global indexes. The former track securities

whose issuers are located in a given region – either geographically or based on the classification of

countries in frontier, emerging, and developed markets –, while the latter track securities whose

issuers are located in multiple regions. With the rise of financial globalization, international indexes

have gained considerable importance, as they constitute the main benchmark for an increasingly

large number of international investors.

Two of the main and most widely tracked international indexes for local currency–denominated

government debt securities are the World Government Bond Index (WGBI) and J.P. Morgan Gov-

ernment Bond Index Emerging Markets (GBI-EM), which are constructed by Citigroup and J.P.

Morgan, respectively. Both indexes represent key benchmarks for international investors in local

currency sovereign debt markets. However, while the former is a global index which tracks the

returns on sovereign bonds denominated in local currency issued by the governments of both de-

veloped and emerging countries, the latter is a regional index which solely focuses on emerging

countries.31

As of 2016, the assets under management benchmarked against the WGBI were approximately

1.5 trillions U.S. dollars, and those benchmarked against the GBI-EM were approximately 200

billion U.S. dollars. Hence, when index providers change the composition of these two indexes,31As a result, the sovereign debt bonds of some emerging countries are included in both indexes. However, giventhat the average market capitalization of the securities in the WGBI is much larger than that in the GBI-EM, theweight of emerging countries in the former is typically much lower than the weight they have in the latter. Forinstance, Mexican local currency sovereign bonds account for about 10% of the GBI-EM, while they account forless than 1% of the WGBI.

55

many international investors wishing to replicate the index composition rebalance their portfolios

accordingly. Index rebalancings therefore trigger capital flows which, as shown in Pandolfi and

Williams (2019), can have important price effects on the value of the local currency sovereign

bonds involved in the rebalancing.

In this paper, in particular, we focus on some large rebalancings in these two indexes which

are due to the inclusion of the following emerging countries: Colombia, Czech Republic, Mexico,

Nigeria, Romania, and South Africa. During our sample period – which spans from 2010 to 2018 –

Argentina was also included in the GBI-EM (in 2017). However, this event is not included in our

sample as the inclusion was driven by the decision of the Argentinian government of removing the

mandatory 120-days holding period for foreign capital, which was taken the day before the country

inclusion announcement.

A.2 Details on Country Inclusion Events

A.2.1 Colombia

On the 19th of March 2014, J.P. released a communication to investors which announced the

inclusion of five Colombian treasury bonds (named TES) into the GBI-EM family of indexes.

The index inclusion was planned to be implemented gradually between May and September 2014,

bringing Colombian weight in the GBI-EM Global Diversified from 3.2% to an estimated 8% (as

three TES were already included in the index before the inclusion episode). The document released

by J.P. Morgan does not contain a time stamp, so we searched the web for news related to this

event. The first news article we found was published by Reuters at 1:52PM (Eastern Standard

Time), which corresponds to 12:52PM in Colombian time, when Colombian markets were still

open.32 Hence, the first trading day for Colombia coincides with the announcement date and is

set on the 19th of March 2014. As regards the drivers of the inclusion, J.P. Morgan states that

this the “[...] result of improved transparency and accessibility for international investors in the32Source: Reuters. https://www.reuters.com/article/colombia-jpmorgan-debt/j-p-morgan-to-boost-colombia-bond-weighting-peso-up-most-in-6-months-idUSL2N0MG12I20140319 (Retrieved on May 6, 2020).

56

local TES market [...]”. Nevertheless, we could not find any relevant news about policy changes

or changes in the functioning of Colombian sovereign debt market which might overlap with the

inclusion announcement made by J.P. Morgan. The most significant regulatory change affecting

Colombian sovereign debt market before the inclusion episode was a tax cut on foreigners investing

in TES which occurred in January 2013, more than one year before the inclusion announcement.

A.2.2 Czech Republic

On the 22nd of February 2017, J.P. Morgan released a communication to investors which announced

the inclusion of nine local-currency denominated Czech sovereign bonds into the GBI-EM family

of indexes. The index inclusion was planned to be implemented gradually between April and June

2017, bringing Czech weight in the GBI-EM Global Diversified from 0% to an estimated 3.3%. The

document was disseminated at 9:44AM (Eastern Standard Time), which corresponds to 3:44PM

in Czech Republic time, when Czech markets were still open. Hence, the first trading day for

Colombia coincides with the announcement date and is set on the 22nd of February 2017. The

inclusion of Czech Republic in the J.P. Morgan GBI-EM was due to reclassification of the country

from a developed to an emerging market, since: “Czech Republic’s GNI per-capita levels falling

below the Index Income Ceiling for three consecutive years”. As a result, Czech sovereign bonds

were excluded from developed markets indexes and included in the Emerging Markets ones. As the

weight of Czech Republic in the index for developed countries was much smaller than that in the

GBI-EM, the outflows due to the exclusion from the former were going to be much smaller than the

inflows due to the inclusion in the latter. According to experts, the transition should have brought

between 3 and 6 billion U.S. dollar inflows to the country.33

33Source: Pensions&Investments Online. https://www.pionline.com/article/20170428/ONLINE/170429837/j-p-morgan-drops-czech-republic-bonds-to-emerging-markets-indexes-inflows-expected (Retrieved on May 6, 2020).

57

A.2.3 Mexico

On the 31st of March 2010, Citigroup announced that Mexican sovereign bonds were eligible for

inclusion in the WGBI, with an estimated weight equal to 0.65%. We searched the web for news

related to this event. The first news article we found was published by Reuters at 12:41PM (Eastern

Standard Time), which corresponds to 11:41AM in Mexican time, when Mexican markets were still

open.34 Hence, the first trading day for Mexico coincides with the announcement date and is set

on the 31st of March 2010. The eligibility announcement stated that: “If Mexico continues to meet

all WGBI criteria for three consecutive months starting with the April 2010 index profile, it will

become the first Latin American and the 24th government bond market to enter the WGBI. Entry

would be effective October 2010”. The criteria refer to the size and ratings of the Mexican sovereign

bonds, and their accessibility to foreign investors. Several news articles highlighted that Mexico

already fulfilled the size and ratings requirements and that during the past year the government

of Mexico improved the liquidity of the government bond market by issuing 30-year bonds, selling

syndicated debt to foreigners and creating a primary dealers program. As a result, Mexico was

eventually included in the WGBI in October 2010.

A.2.4 Nigeria

On the 14th of August 2012, J.P. Morgan released a communication to investors which announced

the inclusion of local-currency denominated Nigerian sovereign bonds (FGN) maturing in 2014, 2019

and 2022, into the GBI-EM family of indexes. The index inclusion was planned to be implemented

gradually between October and December 2012, bringing Nigerian weight in the GBI-EM Global

Diversified from 0% to an estimated 0.59% (the estimate was later revised to 0.72%). The document

released by J.P. Morgan does not contain a time stamp, so we searched the web for news related

to this event. The first news article we found was published by Reuters on the 15th of August at

12:46PM (Eastern Standard Time), which corresponds to 5:46PM in Nigerian time, when Nigerian34Source: Reuters. https://www.reuters.com/article/mexico-index/update-1-citi-says-mexico-eligible-for-wgbi-bond-index-idUSN3121335820100331 (Retrieved on May 6, 2020).

58

markets had already closed.35 Due to the one-day delay between the communication by J.P. Morgan

and the diffusion of the news, we assume that the latest date is the one in which the information

was effectively received by international investors. In fact, we do observe an increase in transaction

of local currency sovereign bonds on the day following the dissemination of the Reuters’ article.

Hence, we set the first trading day after the announcement on the 16th of August 2012. As regards

the drivers of the inclusion, J.P. Morgan stated that this is the result of the improved liquidity of

the Nigerian sovereign debt market, in large part due to the removal of the mandatory one-year

holding period for foreign capital occurred in June 2011, more than one year before the inclusion

announcement.

A.2.5 Romania

On the 16th of January 2013, J.P. Morgan released a communication to investors which announced

the inclusion of local-currency denominated Romanian sovereign bonds (RON) maturing in 2015,

2016 and 2017, into the GBI-EM family of indexes. The index inclusion was planned to be im-

plemented gradually between March and May 2013, bringing Romanian weight in the GBI-EM

Global Diversified from 0% to an estimated 0.54% (the estimate was later revised to 0.87%). The

document released by J.P. Morgan does not contain a time stamp, so we searched the web for news

related to this event. The first news article we found was published by Reuters on the 16th of

January, according to which J.P. Morgan had announced the inclusion overnight.36 We therefore

set the first trading day after the announcement on the 16th of January 2013. As regards the

drivers of the inclusion, J.P. Morgan stated that this was the result of the improved liquidity of the

Romanian sovereign debt market occurred in the 18 months preceding the announcement.35Source: Reuters. https://in.reuters.com/article/us-nigeria-debt-idINBRE87E0TF20120815 (Retrieved on May 6,2020).

36Source: Reuters. https://www.reuters.com/article/romania-debt/jp-morgan-gives-fresh-impetus-to-romania-debt-idUSL6N0AL7IY20130116 (Retrieved on May 6, 2020).

59

A.2.6 South Africa

On the 16th of April 2012, Citigroup announced that 11 Southern African sovereign bonds were

eligible for inclusion in the WGBI. The document released by Citigroup does not contain a time

stamp, so we searched the web for news related to this event. The first news article we found was

published on the 17th of April by Reuters at 07:05AM (Eastern Standard Time), which corresponds

to 1:05PM in the time of South Africa, when markets were still open.37 Hence, we set the first

trading day after the announcement on the 17th of April 2012. The eligibility announcement

stated that: “If South Africa continues to meet all WGBI criteria with the May and June 2012

profiles, it will become the first African government bond market to be included in the WGBI”.

The criteria refer to the size and ratings of the domestic sovereign bonds, and their accessibility

to foreign investors. South Africa already fulfilled the size and ratings requirements at the time of

the announcement. As for the accessibility to foreign investors, it most likely improved in the year

preceding the inclusion, but we could not find more detailed information.

37Source: Reuters. https://af.reuters.com/article/southAfricaNews/idAFL6E8FH3YH20120417 (Retrieved on May6, 2020).

60

A.3 Additional Figures & Tables

Figure A1: Balance of Payments: Private vs. Public Inflows in Each Country

0

1%

2%

3%

4%

Inflow

s over

GD

P

Private Inflows Public Inflows

Average in the 3 yearsbefore the announcement

Year of theAnnouncement

Colombia

0

1%

2%

3%

4%

5%

Inflow

s over

GD

P

Private Inflows Public Inflows

Average in the 3 yearsbefore the announcement

Year of theAnnouncement

Czech Republic

0

1%

2%

3%

4%

Inflow

s over

GD

P

Private Inflows Public Inflows

Average in the 3 yearsbefore the announcement

Year of theAnnouncement

Mexico

0

1%

2%

3%

4%

Inflow

s over

GD

P

Private Inflows Public Inflows

Average in the 3 yearsbefore the announcement

Year of theAnnouncement

Nigeria

0

1%

2%

3%

4%

Inflow

s over

GD

P

Private Inflows Public Inflows

Average in the 3 yearsbefore the announcement

Year of theAnnouncement

Romania

0

1%

2%

3%

4%

Inflow

s over

GD

P

Private Inflows Public Inflows

Average in the 3 yearsbefore the announcement

Year of theAnnouncement

South Africa

This figure depicts the private and the public net inflows to Colombia, Czech Republic, Mexico, Nigeria, Romania and South Africa in the year of the announcementof each country’s inclusion into the corresponding index vs. the average of public and private inflows to these countries in the three years before the announcementepisodes. Private inflows are the sum of foreign direct investments, portfolio equity net inflows and private debt net inflows. Public inflows are net inflows to thecountries’ sovereign debt markets. Both are in U.S. dollars and are normalized by the GDP of each country. Inflows are reported separately for each of the countries inour sample. Data is from the IMF Balance of Payments Statistics and IMF WEO.

61

Figure A2: Alternative Time Windows

-2

-1

0

1

2

Est

imate

d c

oef

fici

ent

2 3 4 5 6

Days from Announcement Date

Gvt&Fin Tradable 90% C.I.

Gvt&Fin-Tradable

-.1

0

.1

.2

.3

Est

imate

d c

oef

fici

ent

2 3 4 5 6

Days from Announcement Date

EFD 90% C.I.

External Financial Dependance

This figure depicts the estimated coefficients of 1(Govt&Fin), 1(Tradable), and EFD in 5 regressions of the form: CARti = β11(Govt&Fin)i + β21(Tradable)i +β3EFDi + εi, where CARi is the cumulative abnormal return of firm i in the t days following the announcement episode, with t ∈ [2, 6]. CARs are computed using a1-factor model where the only risk factor is the return of the MSCI Emerging Markets Index. 1(Govt&Fin) is a dummy variable which equals 1 for government-relatedand financial companies, 1(Tradable) is a dummy which equals 1 for firms operating in tradable sectors, and EFD is a measure of external financial dependancecomputed at the industry-level. In each regression, observations in the top and the bottom percentile of the country-specific distribution of cumulative abnormal returnsare excluded.

62

Table A1: Robustness Tests

Panel A: No companies with zero returns in (-10,10)Demeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

Financial 0.537 0.703 0.682 0.678(0.451) (0.448) (0.430) (0.474)

Government-Related 0.692** 0.628* 0.672* 0.694*(0.350) (0.362) (0.353) (0.357)

Tradable -0.878*** -0.802** -0.844*** -0.816**(0.313) (0.316) (0.315) (0.332)

External Financial Dependance 0.083 0.084 0.089 0.093(0.063) (0.063) (0.061) (0.064)

Financial – Tradable 1.42*** 1.50*** 1.53*** 1.49***(0.48) (0.47) (0.44) (0.49)

Govt-Related – Tradable 1.57*** 1.43*** 1.52*** 1.51***(0.48) (0.49) (0.48) (0.49)

Observations 565 565 565 565R2 0.03 0.03 0.03 0.03

Panel B: No multi-stock companiesDetrended Returns 1-Factor Model 2-Factor Model 5-Factor Model

Financial 0.444 0.543 0.526 0.545(0.388) (0.387) (0.370) (0.408)

Government-Related 0.703** 0.622* 0.656** 0.676**(0.307) (0.322) (0.313) (0.310)

Tradable -0.715*** -0.668** -0.688*** -0.664**(0.262) (0.265) (0.264) (0.279)

External Financial Dependance 0.067 0.064 0.068 0.073(0.051) (0.051) (0.050) (0.052)

Financial – Tradable 1.16*** 1.21*** 1.21*** 1.21***(0.41) (0.40) (0.38) (0.41)

Govt-Related – Tradable 1.42*** 1.29*** 1.34*** 1.34***(0.40) (0.42) (0.41) (0.41)

Observations 652 652 652 652R2 0.03 0.02 0.03 0.02

Panel C : Without Nigeria and Czech RepublicDemeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

Financial 0.413 0.534 0.519 0.507(0.364) (0.363) (0.348) (0.383)

Government-Related 0.633** 0.566* 0.598* 0.625**(0.305) (0.316) (0.308) (0.310)

Tradable -0.716*** -0.661** -0.686*** -0.661**(0.258) (0.261) (0.260) (0.274)

External Financial Dependance 0.066 0.064 0.069 0.072(0.050) (0.050) (0.049) (0.051)

Financial – Tradable 1.13*** 1.20*** 1.20*** 1.17***(0.39) (0.38) (0.36) (0.39)

Govt-Related – Tradable 1.35*** 1.23*** 1.28*** 1.29***(0.40) (0.41) (0.40) (0.41)

Observations 680 680 680 680R2 0.03 0.02 0.02 0.02

This table reports the OLS coefficients of a regression where the dependent variable is the CAR of each firmin the 2 days following the announcement of each country’s inclusion in the corresponding index. In Panel A,we exclude firms whose price never changes in the 20 trading days around the announcement date. In Panel B,we reduce the weight of firms issuing more than one stock, by considering for each of these companies only theaverage 2-day CAR of the company’s traded securities. In Panel C, we exclude Nigeria and Czech Republic. Inall panels, the explanatory variables are: Financial, which is an indicator variable that is equal to 1 for financialfirms; Government-Related, which is an indicator variable that is equal to 1 for government-related firms; Tradable,which is an indicator variable equal to for firms in tradable industries (according to the classification in Mian andSufi (2014)); and External Financial Dependance, which is a measure of firms’ dependance on external financingsources, computed following Rajan and Zingales (1998). The table reports also the difference between the estimatedcoefficient on Financial and that on Tradable, as well as the difference between the estimated coefficients onGovernment-Related and that on Tradable. CARs in the first column (the demeaned returns) are computed asthe the cumulated differences between the daily returns and the average daily return in the year preceding theannouncement. CARs in columns 2 to 4 are computed using three different factor models. The only risk factorin the 1-factor model is the return of the MSCI Emerging Markets Index; the two risk factors in the 2-factormodel are the return of the MSCI Emerging Markets Index and the return of the MSCI World Index; the 5-factormodel include these two risk factors plus small minus big (SMB), high minus low (HML), and momentum (WML)factors. Observations in the top and the bottom percentile of the country-specific distribution of 2-day cumulativeabnormal returns are excluded. Standard errors in parenthesis are clustered at the country-by-industry level. *,**, ***, denote significance at the 10%, 5%, and 1% confidence level, respectively.

63

Table A2: Main Results with Country FE

2-Day Cumulative Abnormal Returns after Announcement Dates

Demeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

Financial 0.533* 0.595** 0.574** 0.579*

(0.284) (0.279) (0.274) (0.300)

Government-Related 0.576** 0.523** 0.579** 0.641**

(0.241) (0.245) (0.244) (0.248)

Tradable -0.419* -0.394* -0.420* -0.365

(0.226) (0.226) (0.227) (0.248)

External Financial Dependance 0.075* 0.072* 0.076* 0.075*

(0.041) (0.042) (0.041) (0.043)

Country FE Yes Yes Yes Yes

Financial – Tradable 0.95*** 0.99*** 0.99*** 0.94***

(0.30) (0.29) (0.28) (0.31)

Govt-Related – Tradable 0.99*** 0.92*** 1.00*** 1.01***

(0.32) (0.33) (0.33) (0.34)

Number of Countries 6 6 6 6

Observations 857 857 857 857

R2 0.04 0.04 0.04 0.03

This table reports the OLS coefficients of a regression where the dependent variable is the CAR of each firm in the 2 daysfollowing the announcement of each country’s inclusion in the corresponding index. The explanatory variables are: Financial,which is an indicator variable that is equal to 1 for financial firms; Government-Related, which is an indicator variable thatis equal to 1 for government-related firms; Tradable, which is an indicator variable equal to for firms in tradable industries(according to the classification in Mian and Sufi (2014)); and External Financial Dependance, which is a measure of firms’dependance on external financing sources, computed following Rajan and Zingales (1998). All regressions include country fixedeffects. The table reports also the difference between the estimated coefficient on Financial and that on Tradable, as well asthe difference between the estimated coefficients on Government-Related and that on Tradable. CARs in the first column (thedemeaned returns) are computed as the the cumulated differences between the daily returns and the average daily return inthe year preceding the announcement. CARs in columns 2 to 4 are computed using three different factor models. The onlyrisk factor in the 1-factor model is the return of the MSCI Emerging Markets Index; the two risk factors in the 2-factor modelare the return of the MSCI Emerging Markets Index and the return of the MSCI World Index; the 5-factor model include thesetwo risk factors plus small minus big (SMB), high minus low (HML), and momentum (WML) factors. Observations in the topand the bottom percentile of the country-specific distribution of 2-day cumulative abnormal returns are excluded. Standarderrors in parenthesis are clustered at the country-by-industry level. *, **, ***, denote significance at the 10%, 5%, and 1%confidence level, respectively.

64

Table A3: Placebo Tests

Panel A: 2-Day Cumulative Abnormal Returns (2 Days before announcement)Demeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

Financial 0.264 0.254 0.248 0.219(0.250) (0.263) (0.264) (0.258)

Government-Related -0.160 -0.227 -0.237 -0.227(0.232) (0.245) (0.249) (0.242)

Tradable 0.402* 0.343 0.337 0.349(0.225) (0.224) (0.223) (0.226)

External Financial Dependance -0.041 -0.044 -0.046 -0.047(0.038) (0.039) (0.039) (0.039)

Number of Countries 6 6 6 6Observations 857 857 857 857R2 0.01 0.01 0.01 0.01

Panel B: 2-Day Cumulative Abnormal Returns (2 Days before announcement)Demeaned Returns 1-Factor Model 2-Factor Model 5-Factor Model

∆Yield×Govt&Fin 0.003 0.001 0.002 -0.000 0.002 -0.000 0.002 -0.001(0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005)

%∆ExchRate× Tradable 0.118 0.158 0.068 0.110 0.067 0.112 0.068 0.112(0.149) (0.155) (0.150) (0.156) (0.151) (0.156) (0.152) (0.157)

External Financial Dependance -0.048 -0.052 -0.054 -0.053(0.036) (0.037) (0.038) (0.037)

Country FE Yes Yes Yes Yes Yes Yes Yes YesNumber of Countries 6 6 6 6 6 6 6 6Observations 861 857 861 857 861 857 861 857R2 0.01 0.01 0.01 0.01 0.00 0.01 0.01 0.01This table reports the OLS coefficients of a regression where the dependent variable is the CAR of each firm over the interval

[t− 3, t− 2], where t is the first trading day after the announcement of each country’s inclusion in the corresponding index.In Panel A, the explanatory variables are: Financial, which is an indicator variable that is equal to 1 for financial firms;Government-Related, which is an indicator variable that is equal to 1 for government-related firms; Tradable, which is anindicator variable equal to for firms in tradable industries (according to the classification in Mian and Sufi (2014)); andExternal Financial Dependance, which is a measure of firms’ dependance on external financing sources, computed followingRajan and Zingales (1998). In Panel B, the explanatory variables are: ∆Y ield×Govt&Fin, which is the product of indicatorvariable that is equal to 1 for financial and government-related firms, and the 2-day change in the 5-year local currencygovernment bond yield, in basis points; %∆ExchRate×Tradable, which is the product of Tradable and the 2-day percentagechange in the exchange rate (computed as the difference in the log of the exchange rate) in percentage points; ExternalFinancial Dependance; and country fixed effects. CARs in the first column (the demeaned returns) are computed as thethe cumulated differences between the daily returns and the average daily return in the year preceding the announcement.CARs in columns 2 to 4 are computed using three different factor models. The only risk factor in the 1-factor model is thereturn of the MSCI Emerging Markets Index; the two risk factors in the 2-factor model are the return of the MSCI EmergingMarkets Index and the return of the MSCI World Index; the 5-factor model include these two risk factors plus small minusbig (SMB), high minus low (HML), and momentum (WML) factors. Observations in the top and the bottom percentile of thecountry-specific distribution of 2-day cumulative abnormal returns are excluded. Standard errors in parenthesis are clusteredat the country-by-industry level. *, **, ***, denote significance at the 10%, 5%, and 1% confidence level, respectively.

65